National Academies Press: OpenBook

Effect of Smart Growth Policies on Travel Demand (2013)

Chapter: Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation

« Previous: Appendix A - Performance Metrics and Tools
Page 117
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 117
Page 118
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 118
Page 119
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 119
Page 120
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 120
Page 121
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 121
Page 122
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 122
Page 123
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 123
Page 124
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 124
Page 125
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 125
Page 126
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 126
Page 127
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 127
Page 128
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 128
Page 129
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 129
Page 130
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 130
Page 131
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 131
Page 132
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 132
Page 133
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 133
Page 134
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 134
Page 135
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 135
Page 136
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 136
Page 137
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 137
Page 138
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 138
Page 139
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 139
Page 140
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 140
Page 141
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 141
Page 142
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 142
Page 143
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 143
Page 144
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 144
Page 145
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 145
Page 146
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 146
Page 147
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 147
Page 148
Suggested Citation:"Appendix B - Smart Growth Area Planning Tool (SmartGAP) Documentation." National Academies of Sciences, Engineering, and Medicine. 2013. Effect of Smart Growth Policies on Travel Demand. Washington, DC: The National Academies Press. doi: 10.17226/22616.
×
Page 148

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

119 A P P e n D I x B Overview Sources Some of the models contained in SmartGAP were derived from work developed from other sources and brought together in this implementation. The primary sources were identified in the description for each model and include the following: • Greenhouse Gas Statewide Transportation Emissions Planning (GreenSTEP) Model Documentation (Novem- ber 2010) prepared by Brian Gregor from the Oregon Department of Transportation, Transportation Planning Analysis Unit (Gregor 2011). • Freight Activity Microsimulation Estimator (FAME) proj- ect conducted by Amir Samimi, Kouros Mohammadian, and Kazuya Kawamura from the University of Illinois at Chicago for the National Center for Freight, Infrastruc- ture, Research and Education (CFIRE) at the University of Wisconsin–Madison and the Illinois Department of Trans- portation (2010). • Highway Economic Requirements System (HERS) model developed for the FHWA in 2005. • National Transit Profile in the National Transit Database. • U.S. DOT’s National Transportation Statistics (2011). • Texas A&M Transportation Institute’s Annual Urban Mobility Report (2009). The urban form models were developed originally for SmartGAP and estimated from the National Household Travel Survey data. Glossary of Variables Used in the Models Table B.1 presents a glossary of variables used in all the models for reference. These are sorted alphabetically by variable name. Census regions (http://www.eia.gov/emeu/mecs/mecs2002/ census.html) are defined by Census divisions and states (Table B.2), as follows: a Census division is a geographic area consisting of several states defined by the U.S. Department of Commerce, Bureau of the Census. States are grouped into four regions and nine divisions. Area types are defined in the National Household Travel Survey (NHTS) data in the Hthur urban/rural variable in Appendix Q of the 2001 NHTS User’s Guide (http://nhts.ornl .gov/2001/usersguide/UsersGuide.pdf). Density was converted into centiles, that is, the raw numbers (persons per square mile) were translated into a scale from 0 to 99: • “Rural” (centiles 19 and less) based on density. • “Small town” (centiles 20 to 39) based on the density. • Population centers were defined if a route through the 8 neighboring cells could be constructed in which the density of successive cells was decreasing or equal. • Population centers with centiles greater than 79 were des- ignated “urban.” • Other centers were classified as “second cities.” • “Suburban” areas of the population centers were defined, using both the cell density and the cell’s density relative to the population center’s density. household and Firm Models Household Age Models The household age model uses a synthesis process that is common in travel modeling to enumerate a set of household records from county-level estimates of population by age. The households are described in terms of the number of peo- ple in each of six age groups (0–14, 15–19, 20–29, 30–54, 55–64, and 65 plus). The aim of the synthesis process is to capture both the overall characteristics of the population, such as average household size, and also the range of those characteristics, such as the distribution of household sizes. The probability distribution linking the population by age data with household membership is obtained from Public Use Smart Growth Area Planning Tool (SmartGAP) Documentation

120 Table B.1. Variables Used in SmartGAP Models Variable Name Description Age0to14 Number of Persons per Household Age 0–14 Age15to19 Number of Persons per Household Age 15–19 Age15to19:VehPerDrvAgePop Persons age 15–19 interacted with vehicles per driver Age20to29 Number of Persons per Household Age 20–29 Age20to29:LogDen Persons age 20–29 interacted with log of population density Age30to54 Number of Persons per Household Age 30–54 Age30to54:LogDen Persons age 30–54 interacted with log of population density Age30to54:VehPerDrvAgePop Persons age 30–54 interacted with vehicles per driver Age55to64 Number of Persons per Household Age 55–64 Age55to64:LogDen Persons age 55–64 interacted with log of population density Age55to64:VehPerDrvAgePop Persons age 55–64 interacted with vehicles per driver Age65Plus Number of Persons per Household Age 65+ Age65Plus:LogDen Persons age 65+ interacted with log of population density Census_rMidwest Dummy variable if household is in the Midwest region Census_rSouth Dummy variable if household is in the Southern region Census_rWest Dummy variable if household is in the Western region Children_City Children Dummy Variable, Second City Area Type Children_Rural Children Dummy Variable, Rural Area Type Children_Suburban Children Dummy Variable, Suburban Area Type Children_Town Children Dummy Variable, Town Area Type CoupleNoKids_City Couple No Kids Dummy Variable, Second City Area Type CoupleNoKids_Rural Couple No Kids Dummy Variable, Rural Area Type CoupleNoKids_Suburban Couple No Kids Dummy Variable, Suburban Area Type CoupleNoKids_Town Couple No Kids Dummy Variable, Town Area Type DrvAgePop Number of driving age persons Fwylnmicap Freeway lane miles per 1000 persons Hhinc_City Household Income ($1000s), Second City Area Type Hhinc_Rural Household Income ($1000s), Rural Area Type Hhinc_Suburban Household Income ($1000s), Suburban Area Type Hhinc_Town Household Income ($1000s), Town Area Type Hhincttl Total annual household income in dollars Hhincttl:Age15to19 Household income interacted with persons ages 15–19 Hhincttl:Age30to54 Household income interacted with persons ages 30–54 Hhincttl:Age55to64 Household income interacted with persons ages 55–64 Hhincttl:Hhvehcnt Household income interacted with household vehicles Hhincttl:Htppopdn Household income interacted with population density Hhincttl:LogDen Household income interacted with log of population density Hhincttl:LogDvmt Household income interacted with daily VMT Hhincttl:LogSize Household income interacted with log of household size Hhincttl:OnlyElderly Household income interacted with elderly populations Hhincttl:Tranmilescap Household income interacted with transit revenue miles Hhincttl:Urban Household income interacted with urban mixed-use area Hhsize Number of persons per household Hhvehcnt Number of vehicles in the household (continued on next page)

121 Table B.1. Variables Used in SmartGAP Models Variable Name Description Htppopdn Census tract population density in persons per square mile Htppopdn:Fwylnmicap Population density interacted with freeway lane miles Htppopdn:Hhvehcnt Population density interacted with household vehicles Htppopdn:OnlyElderly Population density interacted with elderly populations Htppopdn:Tranmilescap Population density interacted with transit revenue miles Htppopdn:Urban Population density interacted with urban mixed-use area LogDen Natural log of the census tract population density LogDen:LogDvmt Log of population density interacted with log of daily VMT LogDen:LogSize Log of population density interacted with log of household size LogDen:Urban Log of population density interacted with urban mixed-use area LogDvmt Log of daily vehicle miles traveled LogIncome Natural log of annual household income LogSize Log of persons per household LogSize:LogDvmt Log of household size interacted with log of daily VMT LogSize:Urban Log of household size interacted with urban mixed-use area OnlyElderly When all persons in the household are over 65 years old OnlyElderly:Fwylnmicap Elderly populations interacted with freeway lane miles OnlyElderly:Tranmilescap Elderly populations interacted with transit revenue miles OnlyElderly_City Only Elderly Dummy Variable, Second City Area Type OnlyElderly_Rural Only Elderly Dummy Variable, Rural Area Type OnlyElderly_Suburban Only Elderly Dummy Variable, Suburban Area Type OnlyElderly_Town Only Elderly Dummy Variable, Town Area Type PowPerCapInc Average per Capita Income (Power Transform) Singleton_City Singleton Dummy Variable, Second City Area Type Singleton_Rural Singleton Dummy Variable, Rural Area Type Singleton_Suburban Singleton Dummy Variable, Suburban Area Type Singleton_Town Singleton Dummy Variable, Town Area Type Tranmilescap Annual transit revenue miles per person Tranmilescap:Fwylnmicap Transit revenue miles interacted with freeway lane miles Tranmilescap:Urban Transit revenue miles interacted with urban areas Tranmilescap:Urban Transit revenue miles per capita interacting with households in an urban mixed-use area Urban Household is in an urban mixed-use area Urban:Fwylnmicap Urban mixed-use areas interacted with freeway lane miles Urban:LogDen Urban mixed-use area interacted with log of population density Urban:LogDvmt Urban mixed-use area interacted with log of daily VMT VehPerDrvAgePop:Age20to29 Persons age 20–29 interacted with vehicles per driver VehPerDrvAgePop:Age65Plus Persons age 65+ interacted with vehicles per driver ZeroVeh Households with no vehicles Note: Some variables are interacted with other variables to include effects from a combination of these variables. For example, household income is interacted with urban mixed-use areas to show that there will be more zero-vehicle households with one driving-age person in the household in urban mixed-use areas as income increases. (continued)

122 Microdata Sample (PUMS) data. The PUMS data were coded into household types based on the number of people in each of the six age groups. Some simplifications were made to repre- sent only the more common household structures in the PUMS data—which still accounted for 99% of all households in PUMS data—by limiting the number of people in the 0 to 14 age group to a maximum of four and in older age groups to a maximum of two. Households with only people in the 0 to 14 age group were filtered out of the PUMS data. The household type summary was converted to a probability of a person in a given age group being in each specific household type. Since a household often comprises several people, applying the prob- abilities to each age group create multiple different estimate of households by type. Gregor (2011) explains the computational process used in the synthesis process to account for this: “An [iterative proportional fitting] IPF process was used to reconcile the household type estimates and create a consis- tent set of households. The first control for the IPF process is to match the population forecasts by age category. The sec- ond control is to create a consistent forecast of the number of households of each type. Each iteration is comprised of the following steps: 1. Persons of each age group are allocated to households by type by applying the calculated probabilities to the num- ber of persons in each age category. 2. The persons allocated by household type are converted to households by type by dividing persons in each age cate- gory and type by the corresponding persons by age for that household type. For example, 100 persons of age 0–14 allo- cated to household type 2-0-0-2-0-0, implies 50 house- holds of that type. 3. The result of step #2 will be several conflicting estimates of the number of households of each type. The method used to resolve the differences in the estimates is the “mean” method that chooses the average of the estimates. 4. The resolved number of households for each type com- puted in step #3 is multiplied by the corresponding num- ber of persons in each age group to yield an estimate of the number of persons by age group and household type. 5. A new table of household type probabilities for each age group is computed from the step #4 tabulation. 6. The sum of persons by age group is calculated from the results of step #4 and subtracted from the control totals of persons by age group to determine the difference to be reallocated. 7. The person differences are allocated to household types using the probabilities calculated in step #5. These steps are repeated until the difference between the maximum number of households and the resolved number of households computed for every household type is less than 0.1 per cent or until a maximum number of iterations (default 100)” (Gregor 2011, pp. 12–13). Household Income Models The household income model is a regression model that estimates household income based on the number of people in each group in the household size and the average per capita income for the region. The regression model’s coef- ficients were estimated by using Census PUMS data and are shown in Table B.3. The dependent variable is a power transform of income, with an exponent of 0.4, following the observed distribution of the PUMS income data. The aver- age per capita income is also power-transformed with the same exponent. The effect on income of additional house- hold member initially increases with age, peaks in the 30 to 54 age group (where people’s earning power and labor force participation typically peaks), and then declines for the older age groups. Table B.2. Census Regions, Divisions, and States Region Division States Northeast New England Connecticut, Maine, Massachusetts, New Hampshire, Vermont, and Rhode Island Middle Atlantic New Jersey, New York, and Pennsylvania Midwest East North Central Illinois, Indiana, Michigan, Ohio, and Wisconsin West North Central Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, and South Dakota South South Atlantic Delaware, District of Columbia, Florida, Georgia, Maryland, North Carolina, South Carolina, Virginia, and West Virginia East South Central Alabama, Kentucky, Mississippi, and Tennessee West South Central Arkansas, Louisiana, Oklahoma, and Texas West Mountain Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, and Wyoming Pacific Alaska, California, Hawaii, Oregon, and Washington

123 Applying a regression model does not recreate the vari- ability in incomes observed in the data, and therefore a ran- dom variable is added to the model’s predictions (drawn from a standard normal distribution). Figure B.1 shows that, with this term added, the model closely replicated the distri- bution of income observed in the PUMS data (Gregor 2011, pp. 16–20). Firm Size Models In the firm size model, county-level estimates of employ- ment by size of business for each industry are transformed into a set of firm records where each firm is defined by the number of employees in each of eight size categories in the firm (1–19; 20–99; 100–249; 250–499; 500–999; 1,000–2,499; 2,500–4,999; and more than 5,000 employees) and by its industry. The firm size model synthesizes the individual firms by enumerating the county-level summaries. The county- level estimates of employment by size of business and indus- try were obtained from County Business Pattern data (http:// www.census.gov/econ/cbp/) (Samimi et al. 2010). Sources The household age and income models were adapted from the GreenSTEP Model Documentation (November 2010) pre- pared by Brian Gregor from the Oregon Department of Trans- portation, Transportation Planning Analysis Unit (Gregor 2011), and the subsequent Energy and Emissions Reduction Policy Analysis Tool Model Documentation (draft August 2011) prepared by Resource Systems Group for the Federal Highway Administration (Resource Systems Group 2011). Table B.3. Household Income Model Description Coefficients Estimate Average per Capita Income (Power Transform) PowPerCapInc 0.792567 Number of Persons per Household Age 0–14 Age0to14 -1.008610 Number of Persons per Household Age 15–19 Age15to19 0.938870 Number of Persons per Household Age 20–29 Age20to29 7.740331 Number of Persons per Household Age 30–54 Age30to54 15.190270 Number of Persons per Household Age 55–64 Age55to64 13.149690 Number of Persons per Household Age 65+ Age65Plus 8.410674 Figure B.1. Distribution of observed and adjusted modeled household incomes.

124 The firm size model was adapted from the FAME project conducted by Amir Samimi, Kouros Mohammadian, and Kazuya Kawamura from the University of Illinois at Chicago for the CFIRE at the University of Wisconsin–Madison and the Illinois Department of Transportation, and the subsequent application of this model as part of the Tour and Supply Chain Modeling for Freight in Chicago project conducted by Resource Systems Group for the Federal Highway Administration. Urban Form Models Household Allocation to Urban Form The purpose of these models is to allocate synthesized house- holds to different types of urban form. These include the type of area where the household or firm resides (urban core, close-in community, suburban, rural), the population and employment density (persons per square mile) of the Census tract where the household or firm resides, and the urban form characteristics of the Census tract where the household or firm resides (urban mixed-use versus other). The synthe- sized households and firms are placed into 13 place types, defined by four area types: • Urban core—includes high-density commercial develop- ments (primarily). • Close-in community—includes medium-density com- mercial and medium-density residential developments. • Suburban—includes low-density residential areas (pri- marily) and low-density commercial development. • Rural—includes greenfield developments only. And five types of development: • Residential—primarily located in suburban areas, but can also occur in close-in community and urban core areas. • Commercial—located in urban core areas (primarily) but also found in close-in communities and suburban areas, but in lower densities. • Mixed use—found in urban core and close-in community areas (primarily) but can also be found in suburban areas. • Transit-oriented development (TOD). • Greenfields—only occur in rural areas. The 13 place types are derived from three area types (urban core, close-in community, and suburban) and four develop- ment patterns (residential, commercial, mixed-use, and transit- oriented development) plus the rural with greenfields place type. The household allocation model comprises the following elements: • Area-type model—a multinomial logit model to predict the probability that a household will reside in each of the area types based on their household income and a set of variables describing the household type. • Model calibration algorithm—an algorithm that adjusts the allocation probabilities so that the overall allocation of households matches the growth by place type input for the scenario. • Area-type allocation—a Monte Carlo simulation to allo- cate each household to a specific area type based on the calibrated probabilities from the previous step. • Development type allocation—a proportional allocation process (based on the development type proportions for the scenario) to allocate households to a development type within each area type. • Population density calculation—a draw from an observed distributions of population densities to assign a specific population tract density to each household, based on their area and development type. Area-Type Model The 2001 NHTS provides a data set that allows the user to identify relationships between demographic data and alloca- tion of households to various area types. A multinomial logit model estimated by using the NHTS data set predicts the probability that a household will reside in each of the area types on the basis of its household income and a set of vari- ables describing the characteristics of the household. The model predicts the area types defined in the NHTS data in the “Hthur” urban/rural variable, a post processed variable that was added to the NHTS data set by Claritas, Inc., and is described in Appendix Q of the 2001 NHTS User’s Guide (http://nhts.ornl.gov/2001/usersguide/UsersGuide .pdf). “The classification that is reflected in the urban/rural variable is based on population density, but not just the den- sity of a specific geography, but the density in context of its surrounding area, or ‘contextual density’. To establish this classification, the United States was divided into a grid to reduce the impact of variation in size (land area) of Census tracts and block groups. Density was converted into centiles, that is, the raw numbers (persons per square mile) were translated into a scale from 0 to 99: • ‘Rural’ (centiles 19 and less) based on the density. • ‘Small town’ (centiles 20 to 39) based on the density. • Population centers were defined if a route through the 8 neighboring cells could be constructed in which the den- sity of successive cells was decreasing or equal. • Population centers with centiles greater than 79 were des- ignated ‘urban.’ • Other centers were classified as ‘second cities.’ • ‘Suburban’ areas of the population centers were defined, using both the cell density and the cell’s density relative to the population center’s density.” (U.S. DOT 2004)

125 At the stage in the overall model process that the area-type model is applied, the population has been synthesized and the household income model has been applied. Therefore, the variables that are available to predict the area type that the household will probably live in are household size, the ages of household members, and household income. In addition, various household structure variables can be constructed to describe the household, such as “singletons” (households that comprise one person of working age). The distributions of these variables were found to be related to the area type where households in the NHTS data set lived. Figure B.2 shows how household size distributions are dif- ferent in each of the five area types defined in the NHTS. Household size skews lowest in the more urbanized area types (second city and urban), and skews highest in the least urban- ized area types (rural and town). Suburban falls in between these two extremes. Figure B.3 shows how the distribution of household income varies across the five area types. The urban area type is notable as having the lowest median income, with the highest median incomes in suburban and town area types, with second city and rural areas in between. Several household structure variables were constructed based on the household size and age variables developed in the household synthesis model. They were developed to seg- ment the household population in to several approximately equal parts (and so are mutually exclusive) based on factors that theoretically affect travel behavior (e.g., presence of children in the household, presence and number of working age adults). The variables are: • Singletons: Households that are made up of one person of working age; • Couple No Kids: Households that are made up of two people of working age; • Children: Households that include children; and • Only Elderly: Households where all household members are 65 years of age or older. Table B.4 shows the variation in area-type distribution among households in the four different area types. The Figure B.2. Distribution of household size for each area type. Figure B.3. Distribution of household income for each area type. Table B.4. Variation in Area-Type Distribution by Household Structure Variable Area Type Singleton (%) Couple No Kids (%) Children (%) Only Elderly (%) Urban 25 19 23 19 Second city 16 21 27 21 Suburban 14 22 28 17 Town 10 23 33 17 Rural 10 25 32 16 Total 100 100 100 100

126 singleton households are the group most heavily skewed toward residence in urban areas. The couple no kids group is relatively evenly distributed by area type, as is the only elderly groups, with the highest proportions in rural and second city area types, respectively. The children group is the group most heavily skewed away from urban areas. The household income variable (specific in thousands of dollars) also follows the trend shown above, with the probability of residence in areas other than urban increasing as income increases, and particularly for suburban and town area types. Table B.5 shows the coefficients on the area-type multi- nomial logit model. The model was estimated by using 19,527 observations, one for each metropolitan area household in the 2001 NHTS (with some screening of data to remove some incomplete records). The model specification includes alternative specific con- stants for four of the five area types, with the urban area type as the base alternative specified without a constant. Member- ship of each of the four household groups is coded as a set of dummy variables of four of the five area types; again, the urban area type is used as the base alternative. The values of the coefficients reflect the trend shown above. For example, the values for singletons are all negative relative to the implicit zero value of urban and the values for children are all positive relative to the implicit zero value of urban. In order to apply the model, the differences between the area types described in the Hthur variable in the NHTS and the area types used in this model must be reconciled. The transla- tion implemented in the application is straightforward: • Urban core = urban. • Close-in community = second city. • Suburban = suburban. • Rural = rural and town. Table B.5. Area-Type Model Description Variable Estimate T-Stat Alternative Specific Constant, Second City Area Type ASC_City -1.07 -13.6 Alternative Specific Constant, Rural Area Type ASC_Rural -1.43 -13.9 Alternative Specific Constant, Suburban Area Type ASC_Suburban -0.348 -5.7 Alternative Specific Constant, Town Area Type ASC_Town -0.903 -13.0 Singleton Dummy Variable, Second City Area Type Singleton_City -0.284 -3.3 Singleton Dummy Variable, Rural Area Type Singleton_Rural -1.07 -8.2 Singleton Dummy Variable, Suburban Area Type Singleton_Suburban -0.505 -7.7 Singleton Dummy Variable, Town Area Type Singleton_Town -0.872 -10.9 Children Dummy Variable, Second City Area Type Children_City 0.119 1.5 Children Dummy Variable, Rural Area Type Children_Rural 0.0962 0.9 Children Dummy Variable, Suburban Area Type Children_Suburban 0.00304 0.1 Children Dummy Variable, Town Area Type Children_Town 0.119 1.8 Couple No Kids Dummy Variable, Second City Area Type CoupleNoKids_City 0.0824 1.0 Couple No Kids Dummy Variable, Rural Area Type CoupleNoKids_Rural 0.0908 0.9 Couple No Kids Dummy Variable, Suburban Area Type CoupleNoKids_Suburban -0.0725 -1.1 Couple No Kids Dummy Variable, Town Area Type CoupleNoKids_Town -0.0918 -1.3 Only Elderly Dummy Variable, Second City Area Type OnlyElderly_City 0.347 4.1 Only Elderly Dummy Variable, Rural Area Type OnlyElderly_Rural -0.347 -2.9 Only Elderly Dummy Variable, Suburban Area Type OnlyElderly_Suburban 0.13 1.9 Only Elderly Dummy Variable, Town Area Type OnlyElderly_Town 0.0623 0.8 Household Income ($1000s), Second City Area Type Hhinc_City 0.00708 9.5 Household Income ($1000s), Rural Area Type Hhinc_Rural 0.00123 1.2 Household Income ($1000s), Suburban Area Type Hhinc_Suburban 0.0123 20.8 Household Income ($1000s), Town Area Type Hhinc_Town 0.0128 19.2 Note: Number of observations = 19,527, number of parameters = 24, initial log likelihood = -31,427.49, final log likelihood = -28,212.36, and rho square = 0.102.

127 The area-type model as estimated will allocate households to the area types in similar overall proportions to those seen in the NHTS sample that was used to estimate the model (with some differences based on for example average income for the scenario). However, it is important for the allocation process to conform to the growth distribution by place type entered as an input to the scenario. This means that the allocation must be adjusted. This is achieved using an iterative calibration pro- cess, during which the alternative specific constants in the model are adjusted until the overall allocation matches the target distribution by place type. During each iteration, the modeled and target area-type shares are compared and the alternative specific constants for each area type are adjusted by a value of natural log (target share/modeled share). Sources The urban form models were developed specifically for this project using place types that were initially developed for the Smart Growth Transect and further refined by the Caltrans Smart Mobility project and combined with place types from Reconnecting America. The models were developed using the NHTS collected by the U.S. DOT. Vehicle Models Vehicle Ownership The vehicle ownership model is a two-stage model that esti- mates the number of vehicles owned by each household in the synthesized population. The first stage of the model allo- cates households to one of four categories based on the ratio of vehicles to driving-age people in the household, using a series of binomial logit models: (a) zero vehicles, (b) fewer than one vehicle per driving-age person, (c) one vehicle per driving-age person, and (d) more than one vehicle per driving- age person. The second part of the model identifies the actual number of vehicles for Category 2 and Category 4 house- holds. The independent variables in the models include free- way supply, transit supply, and urban type variables (Gregor 2011, p. 31). Zero-Vehicle Models Tables B.6 through B.8 show the models for households with zero vehicles, which are segmented into three groups based on the number of driving-age people in the household; that is, one, two, and three or more (Gregor 2011, p. 32). Some variables are interacted with other variables to include effects from a combination of these variables. For example, house- hold income is interacted with urban mixed-use areas to show that there will be more zero-vehicle households with one driving-age person in the household in urban mixed-use areas as income increases. This will counteract the negative coefficient on household income for zero-vehicle households and add to the positive coefficient on households in an urban mixed-use area. It can explain the phenomenon that some higher income households will choose to live in urban mixed- use areas without a car as a lifestyle choice. More Drivers than Vehicles Models The models are segmented into three groups defined by the number of persons of driving age in the household: one driving-age person, two driving-age persons, three or more driving-age persons. Tables B.9 and B.10 show the models for households with more drivers than vehicles. Table B.6. Zero-Vehicle Household Models—1 Driving-Age Person in Household Description Variable Estimate Alternative specific constant (Intercept) -0.683 Total annual household income in dollars Hhincttl -0.00011 Census tract population density in persons per square mi Htppopdn 0.00011 Annual transit revenue miles per person Tranmilescap -0.0362 Household is in an urban mixed-use area Urban 1.03 Household income interacted with population density Hhincttl:Htppopdn 9.06E–10 Household income interacted with transit revenue miles Hhincttl:Tranmilescap 0.00000095 Household income interacted with urban mixed-use area Hhincttl:Urban 0.0000197 Population density interacted with transit revenue miles Htppopdn:Tranmilescap 0.000000963 Population density interacted with urban mixed-use area Htppopdn:Urban -0.0000551 Population density interacted with freeway lane miles Htppopdn:Fwylnmicap -0.000119 Transit revenue miles interacted with freeway lane miles Tranmilescap:Fwylnmicap 0.0577

128 Table B.7. Zero-Vehicle Household Models—2 Driving-Age Persons in Household Description Variable Estimate Alternative specific constant (Intercept) -1.43 Total annual household income in dollars Hhincttl -0.0000679 Household income interacted with population density Hhincttl:Htppopdn 1.42E–09 Household income interacted with elderly populations Hhincttl:OnlyElderly -0.0000355 Population density interacted with transit revenue miles Htppopdn:Tranmilescap 0.00000185 Table B.8. Zero-Vehicle Household Models—3 or More Driving-Age Persons in Household Description Variable Estimate Alternative specific constant (Intercept) -3.49 Total annual household income in dollars Hhincttl -0.000049 Census tract population density in persons per square mi Htppopdn 0.0000972 Household income interacted with population density Hhincttl:Htppopdn 7.31E–10 Transit revenue miles interacted with freeway lane miles Tranmilescap:Fwylnmicap 0.0755 Table B.9. Less than 1 Vehicle per Driving-Age Person Household Models—2 Driving-Age Persons in Household Description Variable Estimate Alternative specific constant (Intercept) -0.263 Total annual household income in dollars Hhincttl -0.0000459 Census tract population density in persons per square mi Htppopdn 0.0000565 When all persons in the household are over 65 years old OnlyElderly 1.74 Household income interacted with population density Hhincttl:Htppopdn 1.19E–09 Household income interacted with transit revenue miles Hhincttl:Tranmilescap 0.000000334 Household income interacted with elderly populations Hhincttl:OnlyElderly 0.00000936 Population density interacted with transit revenue miles Htppopdn:Tranmilescap -0.00000143 Population density interacted with urban mixed-use area Htppopdn:Urban -0.0000475 Population density interacted with elderly populations Htppopdn:OnlyElderly -0.0000271 Transit revenue miles interacted with urban areas Tranmilescap:Urban 0.0295 Elderly populations interacted with transit revenue miles OnlyElderly:Tranmilescap -0.0129

129 Equal Drivers and Vehicles Models The models are segmented into three groups defined by the number of persons of driving age in the household: one driving-age person, two driving-age persons, three or more driving-age persons. Tables B.11 through B.13 show the models for households with one vehicle for each driving- age person in the household. Fewer Drivers than Vehicles Models The models are segmented into three groups defined by the number of persons of driving age in the household: one driving-age person, two driving-age persons, or three or more driving-age persons. Tables B.14 through B.16 show the models for households with more drivers than vehicles. Vehicle Type Models The light truck model predicts the vehicle type—autos or light trucks—for each vehicle in each household. The model is a binary logit model that was estimated using NHTS data. In application, the model is calibrated to match input regional light truck proportions (Gregor 2011, p. 84). Table B.17 shows the model’s coefficients and statistics for the western Census region. “The model includes both a population density and logged population density term. Plots of the relationship between population density and light truck ownership showed there to be a nonlinear relationship. The relation- ship with population density is approximately linear at higher densities while the relationship with the log of population density is approximately linear at lower population densities” (Gregor 2011, p. 85). Table B.10. Less than 1 Vehicle per Driving-Age Person Household Models—3 or More Driving-Age Persons in Household Description Variable Estimate Alternative specific constant (Intercept) 0.934 Total annual household income in dollars Hhincttl -0.0000183 When all persons in the household are over 65 years old OnlyElderly 5.21 Household income interacted with transit revenue miles Hhincttl:Tranmilescap 0.000000166 Household income interacted with urban mixed-use area Hhincttl:Urban 0.0000131 Household income interacted with elderly populations Hhincttl:OnlyElderly -0.00012 Population density interacted with urban mixed-use area Htppopdn:Urban -0.0000489 Population density interacted with transit revenue miles Htppopdn:Fwylnmicap 0.0000893 Urban mixed-use areas interacted with freeway lane miles Urban:Fwylnmicap -0.689 Table B.11. 1 Vehicle per Driving-Age Person Household Models—1 Driving-Age Person in Household Description Variable Estimate Alternative specific constant (Intercept) 0.622 Annual transit revenue miles per person Tranmilescap 0.0233 Household income interacted with population density Hhincttl:Htppopdn 1.13E–09 Household income interacted with transit revenue miles Hhincttl:Tranmilescap -0.000000276 Household income interacted with elderly populations Hhincttl:OnlyElderly 0.0000072 Population density interacted with transit revenue miles Htppopdn:Tranmilescap -0.00000166 Population density interacted with urban mixed-use area Htppopdn:Urban -0.0000454 Population density interacted with transit revenue miles Htppopdn:Fwylnmicap 0.0000408 Elderly populations interacted with transit revenue miles OnlyElderly:Tranmilescap -0.00776

130 Table B.12. 1 Vehicle per Driving-Age Person Household Models—2 Driving-Age Persons in Household Description Variable Estimate Alternative specific constant (Intercept) 0.153 Total annual household income in dollars Hhincttl 0.00000579 Census tract population density in persons per square mile Htppopdn 0.0000402 Household is in an urban mixed-use area Urban -0.381 When all persons in the household are over 65 years old OnlyElderly -0.554 Household income interacted with population density Hhincttl:Htppopdn 2.41E–10 Household income interacted with urban mixed-use area Hhincttl:Urban 0.00000818 Household income interacted with elderly populations Hhincttl:OnlyElderly 0.00000711 Population density interacted with transit revenue miles Htppopdn:Tranmilescap -0.00000179 Population density interacted with urban mixed-use area Htppopdn:Urban -0.0000494 Table B.13. 1 Vehicle per Driving-Age Person Household Models—3 or More Driving-Age Persons in Household Description Variable Estimate Alternative specific constant (Intercept) -1.28 Total annual household income in dollars Hhincttl 0.00000791 Census tract population density in persons per square mile Htppopdn -0.0000576 Household income interacted with population density Hhincttl:Htppopdn 5.38E–10 Transit revenue miles interacted with urban areas Tranmilescap:Urban -0.0204 Table B.14. More than 1 Vehicle per Driving-Age Person Household Models—1 Driving-Age Person in Household Description Variable Estimate Alternative specific constant (Intercept) -1.75 Total annual household income in dollars Hhincttl 0.0000161 Census tract population density in persons per square mile Htppopdn -0.0000567 When all persons in the household are over 65 years old OnlyElderly -1.02 Population density interacted with transit revenue miles Htppopdn:Tranmilescap -0.00000119 Population density interacted with urban mixed-use area Htppopdn:Urban 0.0000453 Urban mixed-use areas interacted with freeway lane miles Urban:Fwylnmicap -0.946 Elderly populations interacted with freeway lane miles OnlyElderly:Fwylnmicap 1.11

131 Table B.15. More than 1 Vehicle per Driving-Age Person Household Models—2 Driving-Age Persons in Household Description Variable Estimate Alternative specific constant (Intercept) -1.96 Total annual household income in dollars Hhincttl 0.00000757 Freeway lane miles per 1,000 persons Fwylnmicap 0.764 When all persons in the household are over 65 years old OnlyElderly -0.665 Household income interacted with population density Hhincttl:Htppopdn 5.78E–10 Population density interacted with transit revenue miles Htppopdn:Tranmilescap -0.00000127 Population density interacted with urban mixed-use area Htppopdn:Urban 0.0000287 Population density interacted with transit revenue miles Htppopdn:Fwylnmicap -0.000156 Transit revenue miles interacted with urban areas Tranmilescap:Urban -0.0227 Table B.16. More than 1 Vehicle per Driving-Age Person Household Models—3 or More Driving-Age Persons in Household Description Variable Estimate Alternative specific constant (Intercept) -1 Census tract population density in persons per square mile Htppopdn -0.000301 Annual transit revenue miles per person Tranmilescap -0.0129 Household income interacted with population density Hhincttl:Htppopdn 2.21E–09 Table B.17. Light Truck Type Model (Western Census Region) Description Variable Estimate Total annual household income in dollars Hhincttl 0.0000106 Number of vehicles in the household Hhvehcnt 0.375 Household is in an urban mixed-use area Urban -3.74 Natural log of the Census tract population density LogDen -0.174 Household income interacted with household vehicles Hhincttl:Hhvehcnt -0.00000377 Population density interacted with household vehicles Htppopdn:Hhvehcnt 0.00000878 Population density interacted with urban mixed-use area Htppopdn:Urban -0.0000549 Urban mixed-use area interacted with log of population density Urban:LogDen 0.445

132 As it is important to match current, past, and forecast light truck proportions, the model calibrates to input light truck proportion for the region by iteratively adding a constant to the model in the application. Vehicle Age Model The vehicle age model assigns an age (vintage) to each vehicle for each household. This allows the model to capture effects such as variations in vehicle age by household income. Higher income households tend to own newer vehicles (Figure B.4), which is important as vehicle age affects fuel economy, and hence fuel expenditures. The model is based on the observed joint and marginal distributions of automobiles and light trucks by age and household income from NHTS data, and is calibrated to match a state’s vehicle age distribution using an IPF procedure (Gregor 2011, p. 87). A Monte Carlo process is used to draw from these joint distributions to select an age for each vehicle (Gregor 2011, p. 88). If the Monte Carlo process is run without a fixed seed, each run will produce different results. Figures B.5 and B.6 show the results of 20 runs of the auto and light truck vehi- cle age model, respectively, for the NHTS Western Census Region survey households. The model runs describe a band Figure B.4. Vehicle age distribution by household income group in Western Census Region households.

133 of results that are consistent with the survey values (Gregor 2011, p. 89). Once each vehicle is identified as an auto or light truck and has an age, it is assigned with the average fuel efficiency for that vehicle type and model year. Fuel efficiencies are mea- sured in gasoline equivalent gallons (i.e., energy content of a gallon of gasoline) and are averaged across fuel types. Model users can vary future fuel economy values. The vehicle model also shares household VMT among a household’s vehicles using a Monte Carlo process to draw from a distribution of annual miles traveled by vehicles in NHTS data (Figure B.7). “The random assignment of mileage proportions to vehicles assumes that households do not optimize the use of their vehicles to minimize fuel use” (Gregor 2011, p. 95). Nonmotorized Vehicle Model The nonmotorized vehicle model predicts the ownership and use of nonmotorized vehicles (where nonmotorized vehicles are bicycles, and also electric bicycles, Segways, and similar vehicles that are small, are lightweight, and can travel at bi cycle speeds or slightly higher than bicycle speeds). According to Gregor (2011), “Modeling the potential future effect of non- motorized vehicles is a challenge because of limited informa- tion about how people will use two-wheeled electric vehicles in U.S. cities and how the use of nonmotorized vehicles in general is affected by the availability of facilities. Given the challenge, the approach taken is to model the potential for diverting house- hold daily vehicle miles traveled (DVMT) to nonmotorized vehicles rather than modeling the use of nonmotorized Figure B.5. Observed and estimated auto age (in years) proportions by income group (20 model runs).

134 vehicles. The core concept of the model is that nonmotorized vehicle usage will primarily be a substitute for short-distance single-occupant vehicle (SOV) travel. Therefore, the core com- ponent of the model is a model of the proportion of the house- hold vehicle travel that occurs in short-distance SOV tours. This model determines the maximum potential for household VMT to be diverted to nonmotorized vehicles given a specified tour length threshold” (p. 107). The factors that determine the total household VMT that is diverted to nonmotorized travel are: 1. The proportion of households that have and use non- motorized vehicles. A model is developed to predict the number of nonmotorized vehicles owned by each house- hold. This model is based on NHTS bicycle ownership data. The model is implemented with a function that allows the user to input an overall nonmotorized vehicle ownership rate for the population. 2. The proportion of SOV tours for which nonmotorized vehi- cles may substitute. A factor is used to include the effect of weather and trip purpose on limiting trips by nonmotorized 0 5 10 15 20 25 30 0. 00 0. 02 0. 04 0. 06 0. 08 0. 10 0. 12 0to20K Observed Estimated 0 5 10 15 20 25 30 0. 00 0. 02 0. 04 0. 06 0. 08 0. 10 0. 12 20Kto40K 0 5 10 15 20 25 30 0. 00 0. 02 0. 04 0. 06 0. 08 0. 10 0. 12 40Kto60K 0 5 10 15 20 25 30 0. 00 0. 02 0. 04 0. 06 0. 08 0. 10 0. 12 60Kto80K 0 5 10 15 20 25 30 0. 00 0. 02 0. 04 0. 06 0. 08 0. 10 0. 12 80Kto100K 0 5 10 15 20 25 30 0. 00 0. 02 0. 04 0. 06 0. 08 0. 10 0. 12 100KPlus Pr op or tio n of lig ht tr uc ks Pr op or tio n of lig ht tr uc ks Pr op or tio n of lig ht tr uc ks Figure B.6. Observed and estimated light truck age (in years) proportions by income group (20 model runs).

135 vehicles. This factor is multiplied by the potential VMT that might be diverted by the household for households having nonmotorized vehicles to calculate the VMT that is diverted. Estimating a Stochastic Model of SOV Travel Proportions The proportion of household VMT in short-distance SOV tours is tabulated from the NHTS day trip data at tour dis- tance thresholds of 5 miles, 10 miles, 15 miles and 20 miles. The data reveals a relationship between the SOV proportions and household income, household size, household VMT, population density, and urban mixed-use character. Fig- ure B.8 shows that the data can be grouped into three catego- ries: (1) households doing no SOV travel, (2) households doing all SOV travel, and (3) households doing some SOV travel, with most households clustered in the first or third 2Veh To ta l N um be r o f S ur ve ye d Ve hi cle s 0.0 0.2 0.4 0.6 0.8 1.0 0 20 0 60 0 10 00 3Veh To ta l N um be r o f S ur ve ye d Ve hi cle s 0.0 0.2 0.4 0.6 0.8 1.0 0 10 0 30 0 50 0 70 0 4Veh (a) (b) (c) (d) To ta l N um be r o f S ur ve ye d Ve hi cle s 0.0 0.2 0.4 0.6 0.8 1.0 0 50 15 0 25 0 35 0 5PlusVeh To ta l N u m b er o f S u rv ey ed V eh ic le s 0.0 0.2 0.4 0.6 Proportion of Household Annual VMT Proportion of Household Annual VMT Proportion of Household Annual VMT Source: NHTS 2001. Proportion of Household Annual VMT 0.8 1.0 0 50 10 0 15 0 Figure B.7. Distribution of proportion of annual vehicle miles traveled by the number of surveyed vehicles in (a) two-vehicle households, (b) three- vehicle households, (c) four-vehicle households, and (d) five-plus-vehicle households. For example, in two-vehicle households (a), the annual house- hold VMT has a normal distribution, where 1,200 surveyed vehicles account for 50% of the annual household VMT, 800 vehicles account for 25% and another 800 vehicles, for about 75%. groups. As the NHTS data represent a single survey day and not averages for the household, stochastic models were esti- mated to predict the proportion of SOV travel that might occur on any given day. These were applied 100 times for each household to derive household averages. Linear models were then estimated by using the household averages; Tables B.18 through B.21 show the coefficients and estimation statistics (Gregor 2011, p. 107). To constrain the results from linear models to be between 0 and 1, a logistic transform was applied to the results, which also improves the model fit. Parameters were estimated for each mileage threshold that maximized the correlation and minimized the difference in the mean values. The form of the logistic function is as follows (Gregor 2011, pp. 118–119): i ( )( ) ( )= + −α −β − −βPropTransform 1 1 exp PropModel 0.5

136 The model application interpolates between the results of the separate distance models, depending on the input tour length threshold. Figure B.9 “shows the distributions in household SOV mileage proportions that result from apply- ing the models with interpolation to a range of thresholds. It also compares the mean values estimated for the 5-, 10-, 15-, and 20-mile thresholds with the mean values from the sur- vey” (Gregor 2011, p. 121). Nonmotorized Vehicle Ownership Model NHTS survey data on the number of full-sized bicycles in the household was used to estimate the nonmotorized vehicle ownership model. Figure B.10 shows how the mean number of full-sized bicycles owned varies with household character- istics and the characteristics of the neighborhood in which the household lives. The linear model predicts the number of bicycles owned by a household based dependent variables including on the ages of household member (AgeXtoY), household income (Hhincttl), household size (Hhsize), the number of vehicles per driving-age household member (Veh- PerDrvAgePop), and the natural log of population density (LogDen). The model’s coefficients are shown in Table B.22. In application, the model calibrates to an input target bicycle ownership level by adjusting the model’s intercept (Gregor 2011, pp. 122–123). Calculating Nonmotorized Weight Vehicle VMT According to Gregor (2011), “Nonmotorized vehicle VMT is calculated as follows:   =LtVehDvmt SovProp PropSuitable LtVehOwnRatio SharingRatio where SovProp = proportion of DVMT traveled by SOV within specified mileage threshold (cal- culated by the SOV proportions model); Table B.18. Estimation Results for Linear Model of the Proportion of Household VMT in SOV Tours Less Than or Equal to 5 Miles Description Variable Estimate Alternative specific constant (Intercept) 0.532 Total annual household income in dollars Hhincttl -0.00000125 Log of Census tract population density in persons per square mi LogDen 0.0192 Log of persons per household LogSize -0.265 Household is in an urban mixed-use area Urban 0.0888 Log of daily vehicle miles traveled LogDvmt -0.122 Household income interacted with daily VMT Hhincttl:LogDvmt 0.000000392 Log of population density interacted with log of daily VMT LogDen:LogDvmt -0.0074 Log of household size interacted with log of daily VMT LogSize:LogDvmt 0.0649 Household income interacted with log of population density Hhincttl:LogDen 4.26E–08 Household income interacted with log of household size Hhincttl:LogSize -0.000000388 Household income interacted with urban mixed-use area Hhincttl:Urban 0.000000295 Log of population density interacted with log of household size LogDen:LogSize 0.00732 Log of population density interacted with urban mixed-use area LogDen:Urban -0.0133 0.0 0.2 0.4 0.6 0.8 1.0 0 5 10 15 20 25 30 SOV Mileage Proportion Pr ob ab ilit y De ns ity Tours <= 5 Miles Tours <= 10 Miles Tours <= 15 Miles Tours <= 20 Miles Figure B.8. Distribution of the proportion of household DVMT in SOV tours. Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx.

137 PropSuitable = proportion of SOV travel suitable for nonmotorized vehicle travel (an input assumption); LtVehOwnRatio = ratio of nonmotorized vehicles to num- ber of driving-age persons (nonmotor- ized vehicle ownership calculated by model); and SharingRatio = ratio of nonmotorized vehicles to driving-age persons necessary for every person to have a nonmotorized vehicle available to meet needs (e.g., a sharing ratio of 0.5 means that one non motorized vehicle could be shared by a two-person household).” Table B.19. Estimation Results for Linear Model of the Proportion of Household VMT in SOV Tours Less Than or Equal to 10 Miles Description Variable Estimate Alternative Specific Constant (Intercept) 0.779 Total annual household income in dollars Hhincttl -0.000000154 Log of Census tract population density in persons per square mi LogDen 0.033 Log of persons per household LogSize -0.359 Household is in an urban mixed-use area Urban 0.332 Log of daily vehicle miles traveled LogDvmt -0.179 Household income interacted with log of daily VMT Hhincttl:LogDvmt 0.000000159 Log of population density interacted with log of daily VMT LogDen:LogDvmt -0.00819 Log of household size interacted with log of daily VMT LogSize:LogDvmt 0.0862 Urban mixed-use area interacted with log of daily VMT Urban:LogDvmt 0.00419 Household income interacted with log of population density Hhincttl:LogDen 1.48E–08 Household income interacted with log of household size Hhincttl:LogSize -0.000000241 Household income interacted with urban mixed-use area Hhincttl:Urban 0.000000366 Log of population density interacted with log of household size LogDen:LogSize 0.00435 Log of population density interacted with urban mixed-use area LogDen:Urban -0.0448 Log of household size interacted with urban mixed-use area LogSize:Urban 0.00509 Table B.20. Estimation Results for Linear Model of the Proportion of Household VMT in SOV Tours Less Than or Equal to 15 Miles Description Variable Estimate Alternative Specific Constant (Intercept) 0.936 Total annual household income in dollars Hhincttl 0.000000701 Log of Census tract population density in persons per square mi LogDen 0.0274 Log of persons per household LogSize -0.366 Household is in an urban mixed-use area Urban 0.339 Log of daily vehicle miles traveled LogDvmt -0.209 Household income interacted with log of daily VMT Hhincttl:LogDvmt -6.51E–08 Log of population density interacted with log of daily VMT LogDen:LogDvmt -0.0051 Log of household size interacted with log of daily VMT LogSize:LogDvmt 0.0857 Urban mixed-use area interacted with log of daily VMT Urban:LogDvmt 0.0152 Household income interacted with urban mixed-use area Hhincttl:Urban 0.000000233 Log of population density interacted with urban mixed-use area LogDen:Urban -0.0503 Log of household size interacted with urban mixed-use area LogSize:Urban 0.0166

138 Sources The vehicle models were adapted from the Greenhouse Gas Statewide Transportation Emissions Planning (GreenSTEP) Model Documentation (November 2010) prepared by Brian Gregor from the Oregon Department of Transportation, Transportation Planning Analysis Unit (Gregor 2011), and the subsequent Energy and Emissions Reduction Policy Ana- lysis Tool Model Documentation (draft August 2011) pre- pared by Resource Systems Group for the Federal Highway Administration (Resource Systems Group 2011). Travel Demand Models Household Vehicle Miles Traveled Models The household vehicle miles travel models estimate average household VMT by first predicting, with a binomial logit model, whether each household travels at all by vehicle on a given day and then calculating, with a linear model, the amount of vehicle travel a household is likely to travel for the day. The models include a stochastic error term to reflect day- to-day variability in household travel. Gregor (2011) describes the model as follows: “As with income, household vehicle travel follows a power distribution. This is shown in the histogram on the left side of Figure B.11. Because the distribution is not normal, transformation is in 0.0 0.2 0.4 0.6 0.8 1.0 0 2 4 6 8 10 12 Proportion of DVMT in SOV Pr ob ab ilit y De ns ity Obs. Mean Est. Mean 5 miles 7.5 miles 10 miles 12.5 miles 15 miles 17.5 miles 20 miles Figure B.9. Comparison of modeled distributions of SOV travel proportions by tour mileage threshold. Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx. Table B.21. Estimation Results for Linear Model of the Proportion of Household VMT in SOV Tours Less Than or Equal to 20 Miles Description Variable Estimate Alternative Specific Constant (Intercept) 1.04 Total annual household income in dollars Hhincttl 0.00000223 Log of Census tract population density in persons per mi2 LogDen 0.0185 Log of persons per household LogSize -0.375 Household is in an urban mixed-use area Urban 0.346 Log of daily vehicle miles traveled LogDvmt -0.224 Household income interacted with log of daily VMT Hhincttl:LogDvmt -0.000000385 Log of population density interacted with log of daily VMT LogDen:LogDvmt -0.000963 Log of household size interacted with log of daily VMT LogSize:LogDvmt 0.0833 Urban mixed-use area interacted with log of daily VMT Urban:LogDvmt 0.0164 Household income interacted with log of population density Hhincttl:LogDen -5.61E–08 Household income interacted with log of household size Hhincttl:LogSize 0.000000215 Household income interacted with urban mixed-use area Hhincttl:Urban 0.000000143 Log of population density interacted with log of household size LogDen:LogSize -0.00277 Log of population density interacted with urban mixed-use area LogDen:Urban -0.0504 Log of household size interacted with urban mixed-use area LogSize:Urban 0.0108

139 1 2 3 4 5 6 Number of Persons in Household 0. 0 1. 0 2. 0 0-20 20-40 40-60 60-80 100+ Household Income ($1000s) 0. 0 1. 0 2. 0 50 300 750 1.5K 3K 7K 17K 30K Neighborhood Density (persons per square mile) 0. 0 1. 0 2. 0 0 1 Urban Mixed-Use Neighborhood (1=yes, 0=no) 0. 0 1. 0 2. 0 0 1 2 3 4 5 6 7+ Number of Vehicles Owned by Household 0. 0 1. 0 2. 0 Figure B.10. Mean number of full-sized bicycles owned per household by household type and environmental characteristics. Table B.22. Household Nonmotorized Vehicle Ownership Model Description Variable Estimate Alternative Specific Constant (Intercept) 0.24 Dummy variable if household is in the Midwest region Census_rMidwest 0.186 Dummy variable if household is in the Southern region Census_rSouth -0.147 Dummy variable if household is in the Western region Census_rWest -0.0152 Number of persons per household Hhsize 0.166 Household income interacted with persons ages 15–19 Hhincttl:Age15to19 0.00000357 Household income interacted with persons ages 30–54 Hhincttl:Age30to54 0.00000249 Household income interacted with persons ages 55–64 Hhincttl:Age55to64 0.00000172 Persons age 15–19 interacted with vehicles per driver Age15to19:VehPerDrvAgePop 0.217 Persons age 20–29 interacted with vehicles per driver VehPerDrvAgePop:Age20to29 0.164 Persons age 30–54 interacted with vehicles per driver Age30to54:VehPerDrvAgePop 0.199 Persons age 55–64 interacted with vehicles per driver Age55to64:VehPerDrvAgePop 0.212 Persons age 65+ interacted with vehicles per driver VehPerDrvAgePop:Age65Plus 0.148 Persons age 20–29 interacted with log of population density Age20to29:LogDen -0.014 Persons age 30–54 interacted with log of population density Age30to54:LogDen -0.0157 Persons age 55–64 interacted with log of population density Age55to64:LogDen -0.0264 Persons age 65+ interacted with log of population density Age65Plus:LogDen -0.0247

140 order to improve the model fit and produce more uniform dis- tribution of residuals. A power transformation with an expo- nent of 0.18 minimizes the skewness of the distribution. This is shown in the right-hand plot. The right-hand plot illustrates why it is necessary to use two models to predict household VMT. The power transform of household VMT places the zero VMT households in a grouping that is discontinuous with the house- holds that have some vehicle travel. Including the zero with the other VMT households would distort the model” (p. 41). Table B.23 shows the coefficients of the zero VMT household model. “The probability of zero VMT increases with higher population density, zero-vehicle ownership, higher levels of transit service, presence of urban mixed-use characteristics, and presence of persons aged 65 or older. The probability of zero VMT decreases with more driving-age persons, higher income, more household vehicles, and more persons in the 30 to 54 age group” (Gregor 2011, p. 43). Table B.24 shows the coefficients of the household VMT model. “Higher incomes, more vehicles, more driving-age persons, and greater freeway supplies are associated with more vehicle travel. Persons age 65 or older, higher popula- tion densities, urban mixed-use characteristics, and higher Figure B.11. Household VMT (left) and power-transformed VMT (right). Table B.23. Zero VMT Household Model Description Variable Estimate Alternative Specific Constant (Intercept) 3.7 Number of driving-age persons DrvAgePop -0.522 Natural log of annual household income LogIncome -0.486 Census tract population density in persons per square mile Htppopdn 0.0000298 Number of persons 65 years old or older in the household Age65Plus 0.32 Annual transit revenue miles per capita Tranmilescap 0.00837 Number of household vehicles Hhvehcnt -0.361 Households with no vehicles ZeroVeh 3.43 Transit revenue miles per capita interacting with households in an urban mixed-use area Tranmilescap:Urban 0.0109

141 levels of public transit service are associated with less vehicle travel” (Gregor 2011, p. 44). A similar approach to that used with the household income model is followed to replicate the observed variability in the VMT distribution. A normally distributed random error is added to the model to reproduce the distribution. “The size of this ‘error term’ (standard deviation) was estimated by tak- ing the square root of the difference in the observed and esti- mated variances of the power-transformed VMT. The final value was calibrated by adjusting the estimated value so that the observed and estimated VMT means match” (Gregor 2011, p. 45). Figures B.12 and B.13 show that the addition on the error term brings the modeled distribution of VMT much closer to the observed distribution. The use of error terms also provides a way to calculate annual average VMT, which is important in order to calculate annual household fuel consumption, costs, and emissions. The NHTS, like most household travel surveys, only collects Table B.24. Household VMT Model Description Variable Estimate Alternative Specific Constant (Intercept) 0.781 Number of persons 65 years old or older in the household Age65Plus -0.0718 Natural log of annual household income LogIncome 0.0869 Census tract population density in persons per square mile Htppopdn -0.00000369 Regional ratio of freeway lane miles per 1,000 persons Fwylnmicap 0.0338 Household is in an urban mixed-use area Urban -0.0518 Number of household vehicles Hhvehcnt 0.0609 Number of driving-age persons DrvAgePop 0.0723 Transit revenue miles per capita interacting with households in an urban mixed-use area Htppopdn:Tranmilescap -5.98E–08 Figure B.12. Observed and estimated distributions of power-transformed VMT for metropolitan households. Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx.

142 data for one survey day so it does not report household annual averages. According to Gregor (2011), “Kuhnimhof and Gringmuth, using data from the multiday German Mobility Panel, found that the day-to-day variation in per- sonal travel for an individual was much greater than the vari- ation between persons. (pp. 178–185). They estimated that 70 per cent of all variance in mileage per person per day was intrapersonal (i.e., day-to-day variation in a person’s travel). If this percentage holds true for variation in household VMT, then day-to-day variation in household vehicle travel would account for 80 percent (0.7/0.88) of the unexplained varia- tion in regional household travel that is captured by the cali- brated random error term” (p. 48). Therefore, as day-to-day travel variation is likely to be responsible for most of the unexplained variation in house- hold travel, the travel models were run many times to develop distributions of vehicle travel for each household. The zero VMT and daily household VMT models were run 100 times for each household in the survey data set. This was repeated 30 times and the results averaged for each household. A linear model for predicting the simulated average VMT was then estimated, as the linear model is much faster in appli- cation. Table B.25 shows the coefficients of the model, which are the same as those used in the daily VMT model shown above. “Higher incomes, more vehicles, more drivers, and a greater freeway supply increase the average household VMT. Owning no vehicles, living at higher population density, more public transit service, and living in an urban mixed-use area decrease the average household VMT” (Gregor 2011, p. 49). Vehicle Cost Models No costs are included in any of the household vehicle travel models. The effects of all variable vehicle costs (costs that vary with the amount of vehicle travel rather than with the number of vehicles owned) on travel are handled by a house- hold travel budget model described in this section. It is important that researchers be able to reasonably account for the effects of fuel prices and similar variable costs such as fuel or carbon taxes on the amount of vehicle travel. There is a significant interest in using pricing mechanisms to affect the demand for vehicle travel, so researchers need a model to esti- mate what the effect of pricing might be and how to account for the effect of future fuel price increases on vehicle travel. The budget approach to modeling is based on the perspec- tive that households make their travel decisions within money and time budget constraints. This was fundamental to the work of Yacov Zahavi in the 1970s and early 1980s (Zahavi 1979). Recently, Michael Wegener has referred back to the work of Zahavi and proposed that models need to be based more on budget constraints and less on observed preferences (Wegener 2008). Figure B.13. Observed and estimated distributions of VMT for metropolitan households. Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx.

143 The basic model concept is as follows: • Household spending on gasoline and other variable costs is done within a household transportation budget that is relatively stable. Households shift expenses between trans- portation budget categories as needed. • As long as it is possible for the household to shift expendi- tures among components of the transportation budget, the household response to changes in fuel prices can be inelastic. However, when fuel prices or other variable costs increase to the point where it is no longer possible to shift money from other parts of the transportation budget, the household will necessarily reduce their travel in direct proportion to the cost increase (ceteris paribus). • The transition between inelastic and elastic behavior will not be abrupt unless there is little time for the household to recognize the impact of the cost increases on the budget or respond to the cost increases. If the changes are more gradual, the transition will be less abrupt. Total household expenditures on transportation have remained fairly constant over the 25-year period from 1984 to 2008. Changes in gasoline prices appear to have had little or no effect on the quantity of gasoline consumed. Changes in price also appear to have had little or no effect on house- hold VMT. The shifting of household expenditures among the different transportation expenditure categories has been responsible for the inelasticity in household gasoline con- sumption and household VMT with respect to gasoline price. Although gasoline consumption and VMT have changed little with respect to price over the last 25 years, it would not be wise to assume that this relationship will continue into the future if gasoline prices increase beyond 2008 levels. If the pre- ceding analysis is correct and households do balance out costs within a fixed transportation budget, there will necessarily be adjustments to gasoline consumption if fuel costs rise to high enough levels. At some point, it would no longer be possible to reduce vehicle purchases or other vehicle expenditures in order to avoid reducing gasoline consumption. Vehicles still need to be insured, licensed, maintained, and repaired. Vehicle purchases can be put off, but not indefinitely. When a house- hold reaches the point when it is no longer possible to shift expenditures to other categories, household members will have to reduce gasoline consumption. If they cannot increase the fuel economy of the vehicles they drive, they will have to reduce the amount that they drive. To model the transportation budget it is necessary to esti- mate the size of the transportation budget. Then it is neces- sary to estimate the maximum proportion of that budget that can be used for fuel and other variable costs. The budget model is very simple. First, a base level of travel is estimated using the average household VMT model described in the previous section. This model estimates household travel as a function of the household income, number and ages of persons in the household, population density and mixed-use character where the household resides, freeway supply, and public transit supply. Because 2001 is at the end of a long period of low fuel prices, the model reflects an equilibrium Table B.25. Regional Household Average VMT Model Description Variable Estimate Alternative Specific Constant (Intercept) 0.647 Households in the Census Midwest region Census_rMidwest 0.0000717 Households in the Census South region Census_rSouth -0.000735 Households in the Census West region Census_rWest 0.00155 Natural log of annual household income LogIncome 0.107 Census tract population density in persons per square mile Htppopdn -0.00000316 Number of household vehicles Hhvehcnt 0.058 Households with zero vehicles ZeroVeh -0.59 Annual transit revenue miles per capita Tranmilescap -0.000176 Regional ratio of freeway lane miles per 1,000 persons Fwylnmicap 0.0337 Number of driving-age persons DrvAgePop 0.0857 Number of persons 65 years old or older in the household Age65Plus -0.0768 Household is in an urban mixed-use area Urban -0.0613 Transit revenue miles per capita interacting with households in an urban mixed-use area Htppopdn:Tranmilescap -0.000000115

144 condition between low fuel prices and other factors affecting vehicle travel. It therefore is a good representation of a base level of vehicle travel without budget constraints. Second, a maximum household budget expenditure is cal- culated based on the assumption about the maximum pro- portion of household income that may be spent (a default of 10% of household income is assumed, but the model is not hard-coded with this default value). It is possible to input other values. The most recent consumer expenditure survey (2010) has a 12% transportation expenditure (http://www.bls.gov/ opub/focus/volume2_number12/cex_2_12.htm). From this budget and the base forecast of vehicle travel, a threshold level for average household cost per mile of travel is calculated. If the cost per mile is less than the threshold level, then the household can continue to travel at the base level. If the cost per mile is greater than the threshold, then the household has to reduce the amount of travel in proportion to the increase in cost above the threshold. Figure B.14 shows the shape of the curve for hypothetical households having different incomes. The flat portions of the curves show the potentially inelastic portions to the left of the threshold. The perfectly elastic por- tions of the curves are to the right of the cost thresholds. The figure also shows transition curves that may be speci- fied between the inelastic and elastic portions of the curves. The transition curves are calculated by using a hyperbolic cosine function that is symmetrical about the average cost threshold. These transition curves are specified by the location of the start of the transition between the base cost per mile and the threshold cost per mile. Several tests were run on this budget model. The purpose of the first set of tests was to calculate the elasticity of travel demand with respect to fuel price. The VMT models were applied to the respective household data sets over a range of fuel prices from $1 to $10 per gallon. Fuel price elasticities were then calculated at each dollar increment in the range. Table B.26 shows the results of modeling assuming a full tran- sition. Elasticities increase as prices increase. They decrease as incomes increase. This appears to be reasonable behavior consistent with the budget principle. The low elasticities at low price increases are consistent with other studies that have found recent price elasticities to be low. The household budget approach solves the problems exhibited by previous models. It matches recent travel trends that have exhibited low fuel price elasticity. It also is sensitive to large increases in prices. Moreover, it does this with a sim- ple and strong conceptual model. Bus and Passenger Rail Vehicle Miles Traveled Annual transit revenue miles are calculated to provide inputs to the household vehicle ownership and travel models. It is a straightforward process to compute total bus and passenger rail vehicle miles traveled by multiplying the revenue miles by 0 5 10 15 0 10 20 30 40 50 60 70 Dollars Per Gallon Av er ag e Ho us eh ol d DV M T $20,000 Income $40,000 Income $60,000 Income $80,000 Income No Transition 50% Transition 100% Transition Base Price Figure B.14. Illustration of budget functions and transition curves. Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx. Table B.26. Fuel Price Elasticity Calculated from Application of Regional VMT Model and Budget Model Fuel Price Range (Dollars per Gallon) Income ($) $1–$2 $2–$3 $3–$4 $4–$5 $5–$6 $6–$7 $7–$8 $8–$9 $9–$10 0–30,000 -0.062 -0.288 -0.495 -0.658 -0.776 -0.854 -0.905 -0.939 -0.960 30,000–40,000 -0.021 -0.150 -0.321 -0.482 -0.619 -0.726 -0.804 -0.860 -0.899 40,000–50,000 -0.016 -0.117 -0.268 -0.428 -0.561 -0.669 -0.754 -0.816 -0.862 50,000–70,000 -0.006 -0.068 -0.198 -0.355 -0.498 -0.619 -0.711 -0.781 -0.834 More than 70,000 -0.002 -0.032 -0.102 -0.201 -0.315 -0.430 -0.538 -0.629 -0.704

145 a factor that accounts for nonrevenue service travel. An aver- age of 1.12 is used. Fleet average bus fuel economy and rail energy efficiency are calculated similarly to the way in it is calculated for light vehicles. Bus and rail fuel economy by model year is an input to the model. Different assumptions on future improvements to fuel economy can be modeled by varying these inputs. Buses and rail cars are assigned to age bins based on a refer- ence age distribution and input assumption for adjusting the 95th percentile vehicle age. The age proportions by model year are used with the fuel economy inputs by model year to compute an overall fleet average fuel economy. Heavy Truck VMT Model The forecast of heavy truck VMT is straightforward. Future total regional income is calculated from the forecasts of pop- ulation and average per capita income. Then the percentage change in total regional income from the base year is calcu- lated. The base year heavy truck VMT is multiplied by this change and any relative change factor the user may have sup- plied. The Federal Highway Cost Allocation Study is used to calculate the average proportion of truck VMT by urban area functional class (Table B.27). Average fleet fuel economy for heavy trucks is calculated similarly to the way in it is calculated for light vehicles. Heavy truck fuel economy by model year is an input to the model. Different assumptions on future improvements to fuel econ- omy can be modeled by varying these inputs. Heavy trucks are assigned to age bins based on a reference truck age distri- bution and input assumption for adjusting the 95th percen- tile truck age. The age proportions by model year are used with the fuel economy inputs by model year to compute an overall fleet average fuel economy. Sources The vehicles models were adapted from the Greenhouse Gas Statewide Transportation Emissions Planning (GreenSTEP) Model Documentation (November 2010) prepared by Brian Gregor from the Oregon Department of Transportation, Transportation Planning Analysis Unit (Gregor 2011), and the subsequent Energy and Emissions Reduction Policy Analy- sis Tool Model Documentation (draft August 2011) prepared by Resource Systems Group for the Federal Highway Admin- istration (Resource Systems Group 2011). Congestion by Functional Class The congestion model estimates speed and hence delay and the impact on fuel economy of congestion for freeways and arteri- als and for light vehicle, trucks, and buses. The first step of the model allocates VMT to a simplified functional class break- down of freeways, arterials, and other roads. For trucks and buses, VMT is allocated using fixed proportions (as described above). The auto and light truck proportion on freeways and arterials versus other roads is first calculated using a fixed pro- portion from the Federal Highway Cost Allocation Study. Then auto and light truck VMT is allocated between freeways and arterials using this regression model, estimated using data from the 2009 Texas A&M Transportation Institute’s Urban Mobility Report (based on 2007 data) augmented with VMT proportions calculated from Highway Statistics Table HM-71:  Freeway VMT Proportion 0.07686 2.59032 Freeway Lane Mile Ratio = + Freeway lane mile ratio is the lane miles of freeways divided by the sum of the lane miles of freeways and arterials. When the ratio is applied to the VMT reported in the 2009 version of the Urban Mobility Report, the relationship is linear (Figure B.15). The next stage of the congestion model predicts the pro- portions of VMT experiencing different levels of congestion using models estimated from Urban Mobility Report catego- ries and data. The level of congestion is described using five categories: uncongested, moderately congested, heavily con- gested, severely congested, and extremely congested. Fig- ure B.16 shows the relationship between the traffic volume per lane and the amount of VMT allocated to each congestion cat- egory for freeways; similar relationships are used for arterials. The portion of allocated VMT is calculated the four categories shown, with the proportion for the moderately congested category calculated as the remainder (Gregor 2011, p. 131). Speeds by Congestion Levels The relationship between the congestion category and speeds is based on the Urban Mobility Report, which provides an average trip speed for each congestion level and allows VMT to be allocated to speed bins. Then fuel economy is calculated Table B.27. Heavy Truck VMT Proportions by Urban Functional Class Functional Class Heavy Truck Proportion (%) Principal Arterial—Interstate 8.3 Principal Arterial—Other Freeway or Expressway 5.6 Principal Arterial—Other 5.4 Minor Arterial 4.2 Collector 3.8 Local 3.6

146 Figure B.15. Relationship of freeway to arterial VMT. Figure B.16. Freeway VMT percentages by congestion level versus average daily traffic per lane.

147 by using speed and fuel economy curves, shown in Figure B.17. Two sources are used for these curves: those compiled by the FHWA using the EPA’s MOVES model (Jeff Houk, Federal Highway Administration, personal communication with Brian Gregor, the Oregon DOT) and from the Transportation Energy Data Book (Davis et al., Table 4.29). The fuel economy values are indexed to fuel economy values at 60 mph. The default values used in the model are the curves prepared by Jeff Houk for buses and trucks and those based on the Energy Data Book for light vehicles (Gregor 2011, p. 136). The speed and fuel economy curves are normalized for used in the model. According to Gregor (2011), “Normalization was simply the division of the fuel economy at each speed level by the fuel economy at the assumed free flow speed for each func- tional classification (freeway = 60 MPH, arterial = 30 MPH, other = 20 MPH). This normalization is necessary because average fleet fuel economy values already account for the split of travel between ‘highway’ and ‘city’ driving. If fuel economy were adjusted relative to freeway speeds, there would be a dou- ble counting of the effects of ‘city’ driving on fuel economy. Bus fuel economy normalization on arterials and other roadways is based on the respective average estimated service speeds, 20 MPH and 15 MPH, respectively. Figure [B.18] shows the normalized curves for freeways. Figure [B.19] shows the nor- malized curves for arterials. In Figure [B.19] the bus value is 1 at 20 MPH rather than 30 MPH. That is because the assumed route speed for buses on arterials is 20 MPH. The model caps bus speeds at 20 MPH on arterials. Since it is assumed that ‘other roadways’ are unaffected by congestion, fuel economy for VMT occurring on these roadways is not adjusted in response to speed” (p. 137). Figure B.17. Comparison of fuel economy–speed curves from Houk ( personal communication) and the Transportation Energy Data Book (Davis et al. 2010). Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx. Sources The congestion models were adapted from the Greenhouse Gas Statewide Transportation Emissions Planning ( GreenSTEP) Model Documentation (November 2010) prepared by Brian Gregor from the Oregon Department of Transportation, Transportation Planning Analysis Unit (Gregor 2011), and the subsequent Energy and Emissions Reduction Policy Ana- lysis Tool Model Documentation (draft August 2011) pre- pared by Resource Systems Group for the Federal Highway Source: Gregor 2011. Figure B.18. Freeway speed and fuel economy relationships by vehicle type. Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx.

148 Administration (Resource Systems Group 2011). As part of the model development and validation process, GreenSTEP evaluated data from the 2009 Urban Mobility Report pre- pared by the Texas A&M Transportation Institute to deter- mine the relationship between freeway and arterial lane miles (Texas A&M Transportation Institute 2009). The Green- STEP model development process also evaluated this same report to identify the relationship between VMT by freeways and arterials with the resulting level of congestion. Policies Parking Pricing Policies Parking pricing is a trip-based cost, commonly paid for at one or both ends of a trip, and sometimes paid for on a monthly basis. The standard practice for handling parking pricing in urban travel demand models is to include it in the trip costs for auto travel. That is what is done here, but in a more gen- eral way. Two types of parking costs are addressed in the model: parking costs at places of employment and parking costs at other places. Daily parking costs are calculated for each household and added in with other variable costs. For employer-based parking, the proportion of employees that pay for parking is a policy input. Employer-based parking includes parking provided at the employment site as well as parking in other parking facilities near the employment site. A related policy variable is the availability of free parking in the vicinity of employment sites. This is specified as the ratio of employment parking to available parking in the vicinity of employment sites. It is assumed that the proportion of employees who pay for parking is a function of the proportion of employers who charge for parking and the employment parking proportion of total parking available in the vicinity of employment sites. After the proportion of workers paying for parking has been calculated, the proportion of working age adults paying for parking is calculated by using the labor force participation rate (0.65). Another policy input is the proportion of employment parking that is converted from being free to payment under a “cash-out buy-back” type of program. Under these programs all employees are charged for employer-provided parking but they are also provided with a stipend equal to the parking cost regardless of whether they use the parking or not. This pro- vides an incentive for employees to carpool or use other modes of transportation to get to work. The rate per working age adult and the proportion of cash- out buy-back parking are used in a Monte Carlo process to determine the number of adults in the household who have to pay for parking at their place of work and the number who pay through a cash-out buy-back program. Households are charged the daily parking rate for the number of working age persons identified as paying for parking. Their income is increased for the number of working age persons identified as participating in cash-out buy-back programs with the amount equal to the daily parking rate times the number of working days in a year (260). Parking charges associated with nonwork travel are specified in terms of the proportion of nonwork vehicle trips that incur parking costs. The daily household parking cost for nonwork travel is calculated as the proportion of nonwork trips that incur a parking cost times the average proportion of VMT that is for nonwork travel (0.78) times the average daily parking. The parking pricing model is adapted from the Greenhouse Gas Statewide Transportation Emissions Planning (Green- STEP) Model Documentation (November 2010) prepared by Brian Gregor from the Oregon Department of Transporta- tion, Transportation Planning Analysis Unit (Gregor 2011), and the subsequent Energy and Emissions Reduction Policy Analysis Tool Model Documentation (draft August 2011) pre- pared by Resource Systems Group for the Federal Highway Administration (Resource Systems Group 2011). ITS Policies The intelligent transportation system (ITS) policy measures the effects of incident management supported by ITS. The congestion model contains two sets of relationships between congestion and speed, derived from the Urban Mobility Report. One is with incidents and one is without incidents. According to Gregor (2011), “The model uses the mean speeds with and without incidents to compute an overall Source: Gregor 2011. Figure B.19. Arterial speed and fuel economy relationships by vehicle type. Color version of this figure: www.trb.org/Main/Blurbs/168761.aspx.

149 Source: Gregor 2011. Figure B.20. Estimated freeway speeds by congestion level (upper line in each graph, no incidents; lower lines, with incidents). Color versions of the figure: www.trb.org/Main/Blurbs/168761.aspx. Source: Gregor 2011. Figure B.21. Estimated arterial speeds by congestion level.

150 average speed by road type and congestion level, as shown in Figure B.20 for freeways and Figure B.21 for arterials. The approach provides a simple level of sensitivity testing of the potential effects of incident management programs on emis- sions. An average speed is calculated for each congestion level by interpolating between the incident and non-incident speeds based on an assumed reduction in incidents. For example, an assumed reduction of 0.5 would result in a cal- culated value that is midway between the incident and non- incident speed levels. Speeds are treated differently for autos, light trucks, and heavy trucks than for buses. For the former, speeds are derived from the congestion models just described for freeways and arterials. Speeds on other roadways are assumed to be 20 MPH and unaffected by congestion. For bus VMT on freeways, speeds are those calculated for freeways as described, but for arterials and other local streets, speeds are based on bus service characteristics derived from transit agency data. The assumed speed for arterial service is one standard deviation above the mean of all bus routes (21 MPH). The assumed speed for other roadway service is one standard deviation below the mean (13 MPH). These values are rounded to 20 MPH and 15 MPH, respectively” (pp. 135–136). The approach to estimating the effects of ITS programs is adapted from the Greenhouse Gas Statewide Transporta- tion Emissions Planning (GreenSTEP) Model Documen- tation (November 2010) prepared by Brian Gregor from the Oregon Department of Transportation, Transportation Planning Analysis Unit (Gregor 2011), and the subsequent Energy and Emissions Reduction Policy Analysis Tool Model Documentation (draft August 2011) prepared by Resource Systems Group for the Federal Highway Administration (Resource Systems Group 2011). References Davis, S. C., S. W. Diegel, and R. S. Boundy. 2010. Transportation Energy Data Book, 29th ed., U.S. Department of Energy, Oak Ridge National Laboratory. Gregor, B. 2011. Greenhouse Gas Statewide Transportation Emissions Plan- ning (GreenSTEP) Model Documentation. Oregon Department of Transportation, Transportation Planning Analysis Unit, Portland. Highway Economic Requirements System (HERS) Model. 2005. FHWA, U.S. Department of Transportation. National Household Travel Survey User’s Guide. 2001. Appendix Q. http://nhts.ornl.gov/2001/usersguide/UsersGuide.pdf. National Transit Profile in the National Transit Database. http://www .ntdprogram.gov/ntdprogram/data.htm. Accessed May 29, 2014. Resource Systems Group. 2011. Federal Highway Administration Energy and Emissions Reduction Policy Analysis Tool. FHWA, U.S. Depart- ment of Transportation: Washington, D.C. Samimi, A., K. Mohammadian, and K. Kawamura. 2010. A behavioral freight movement microsimulation model: Method and data. Transportation Letters: The International Journal of Transportation Research 2: 53–62. Texas A&M Transportation Institute. 2009. Annual Urban Mobility Report. Texas A&M State University, College Station. U.S. DOT. 2004. 2001 National Household Travel Survey User’s Guide. (Version 3). U.S. Department of Transportation, Washington D.C. Available at http://nhts.ornl.gov/2001/usersguide/UsersGuide.pdf. U.S. Department of Transportation. 2011. National Transportation Statistics. Wegener, M. 2008. After the Oil Age: Do We Need to Rebuild our Cities? Presentation at the 5th Oregon Symposium on Integrating Land Use and Transport Models, Portland, Ore. Zahavi, Y. 1979. The UMOT Project. UDOT, Research and Special Pro- grams Administration, Washington, D.C. Available at http://www .surveyarchive.org/Zahavi/UMOT_79.pdf.

Next: Capacity Technical Coordinating Committee »
Effect of Smart Growth Policies on Travel Demand Get This Book
×
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

TRB’s second Strategic Highway Research Program (SHRP 2) Report S2-C16-RR-1: Effect of Smart Growth Policies on Travel Demand explores the underlying relationships among households, firms, and travel demand. The report also describes a regional scenario planning tool that can be used to evaluate the impacts of various smart growth policies.

SHRP 2 Capacity Project C16 has also released the SmartGAP User’s Guide. SmartGAP is a scenario planning software tool that synthesizes households and firms in a region and determines their travel demand characteristics based on their built environment and transportation policies.

A zipped version of the SmartGAP software is available for download.

Software Disclaimer - SmartGAP is offered as is, without warranty or promise of support of any kind either expressed or implied. Under no circumstance will the National Academy of Sciences or the Transportation Research Board (collectively "TRB") be liable for any loss or damage caused by the installation or operation of this product. TRB makes no representation or warranty of any kind, expressed or implied, in fact or in law, including without limitation, the warranty of merchantability or the warranty of fitness for a particular purpose, and shall not in any case be liable for any consequential or special damages.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!