Automobile Ownership Model Prepared by: The National Center for Smart Growth Research and Education at the University of Maryland* Cinzia Cirillo, PhD, March 2010 *The views expressed do not necessarily represent those of the University of Maryland, the Maryland Department of Transportation, or the State of Maryland.
Automobile Ownership Model Month 2010 EXECUTIVE SUMMARY In this task we have developed a car ownership model based on Maryland respondents to the 2001 Nationwide Household Travel Survey (NHTS). Using the NHTS data initial studies were conducted to determine the most critical relationships between household characteristics and auto ownership. Following this a logit model was developed which can forecast whether a household would have 0, 1, 2,, 4 or more than four automobiles. Variables in the model include household income, size, number of children, number of workers, number of licensed drivers, education level of the household head, type of location ranging from rural to urban, density and whether the household owns a home or rents. After developing the model, sensitivity tests were conducted which reflects the model s ability to respond to future policies. The sensitivity tests indicated that when density increased auto ownership declined and when unemployment increased auto ownership declined. The model may be used in a stand alone mode to forecast revenues from automobile sales taxes and registration fees or combined with the statewide transportation model to analyze the interactive effects among traffic congestion, land use, trip making and auto ownership.
Automobile Ownership Model NCSGRE March 2010 Page 2 1 BACKGROUND The National Center for Smart Growth Research and Education (NSCG) at the University of Maryland signed a Memorandum of Understanding with the Maryland Department of Transportation (MDOT) to work jointly on research related to transportation in Maryland. The work plan for 2009, the first year of the contract, contained six research tasks. This report is NCSG s submission for Tasks A1, Car Ownership Model. The original statement of scope for this task was: This task will result in modeling framework to estimate car ownership (number and type of vehicles) at the household level for Maryland residents. The deliverable will include documentation on the development and estimation of this model(s), use for forecasting automobile purchasing and revenues related to auto ownership, and integration into the MSTM. Task A1 calls for the development of a model which will produce the tools needed to understand and predict consumer preferences on vehicle ownership, as a function of socio- demographic, economic transportation and land development characteristics. We have successfully developed a tool which estimates the impact of changes in density, type of land use and type of housing on car ownership. When combined with the statewide model and the land use model we will be able to analyze the interrelated effects between congestion, land use, and car ownership. It can also be used in a stand alone manner to forecast revenues related to vehicle sales and registration. The aim of this report is to present the vehicle ownership model developed for the State of Maryland. The rest of this report has five sections: Section 2, Introduction Section, Data Description, brief quantitative description of the data sources and main determinants in car ownership Section 4, Methodology, methodologies used to model the problem Section 5, Modeling Framework, the vehicle ownership model formulation, estimation, validation and application
Automobile Ownership Model NCSGRE March 2010 Page Section 6, Conclusions, conclusions and future challenges. 2 INTRODUCTION Vehicle ownership modeling is being used for a wide variety of purposes. Land use researchers and planners implement vehicle ownership models into trip generation for more accurate planning forecasting. State Departments of Transportation and State Environmental departments develop vehicle ownership and vehicle use models to forecast transport demand, energy consumption and emission levels, as well as the likely impact of policy measures. States may also use car ownership models to estimate revenues from sales taxes or auto registration fees. Car manufacturers apply models to the consumer valuation of attributes relative to cars that are not yet on the market. Oil companies want to predict the future demand for their products and might benefit from car ownership models. Over the last few decades there has been a great increase in the number of cars in the United States. Also there has been a steady increase in the average number of cars per household and a consistent increase in the proportion of households with access to more than one vehicle. Given the importance of vehicle ownership on both transport and land-use planning and its relationship with energy consumption, the environment and health, the growth in the number of vehicles and their use has been one of the most intensely researched transport topics over many years. DATA DESCRIPTION The car ownership model has been developed using data from the 2001 Nationwide Household Travel Survey (NHTS) conducted by the Federal Highway Administration (FHWA). The NHTS collected travel data from a national sample of the civilian, non-institutionalized population of the United States. There are approximately 70,000 households in the final 2001 NHTS dataset while 4,240 households (totally 7012 vehicles) in Maryland area. The total number of valid observations used in the models is about 20 (households).
Automobile Ownership Model NCSGRE March 2010 Page 4 The NHTS was conducted as a telephone survey, using Computer-Assisted Telephone Interviewing (CATI) technology. The 2001 NHTS data set includes the information that is needed in the model, but is not limited to: Household data on the relationship between household members, education level, income, housing characteristics, and other sociodemographic information; Information on each household vehicle, including year, make, model, and estimates of annual miles traveled; Data about drivers, including information on travel as part of work. A systematic examination of the data was undertaken and the following set of explanatory variables identified: household income, number of adults, number of workers, number of licensed drivers, household structure, education of household head, household location. Each is discussed in turn below. It is important to note that the variables included within the analysis are simply those contained within the published datasets referred to above. Although these variables are commonly used for car ownership models [1] it must be acknowledged that the household car ownership decision is clearly influenced by innumerable factors and arguably more detailed information could be gained through a data collection exercise involving detailed in-depth interviews. Household income in often used to explain car ownership. Not surprisingly, vehicle ownership increases directly with income. As shown in Figure 1, 67.14 percent of households with annual incomes under $5,000 own no vehicles, Figure 1 Percent of Households Owning One or More Vehicles by Annual Household Income Percent of Households 100% 90% 80% 70% 60% 50% 40% 0% 2.14 0.00 0.00 1.47 0.00 1.88 0.00 2.01 0.00 1.11 0.00 1.95 2.4 0.00 2.47 0.00 0.18 0.55 0.00 0.58 0.00 0.00 1.9 4.77 2. 1.06 0.4 0.00 0.6 0.44 1.74.40 2.22 0.00 0.66 0.00 0.00 0.89 0.00 2.46 2.77 0.52 0.52 0.00 1.98 0.78 0.00 0.22 1.97 0.19 0.25 0.8 1.54 1.5 1.5 8.57 8.82 5.97 6.01 4.52 10.80 4.19 12.44 6.94 6.99 1.4 18.97 9.54 14.92 15.12 17.2 10.4 22.92 29.5 18.94 20.7 22.14 27.65.58 22.67 22.60 2.17 22.26 26.97 9.44 4.50 7.64 40.70 41.91 54.29 49.4 45.40 60.87 56.66 50.67 5.24 45.15 50.18 57.92 67.14 62.06 50.00 42.57 9 8 7 6 5 4 2 1 0 20% 47.89 41.8 4.0 1.06 10% 20.61 12.87 11.66 7.89 5.2 4.89 0% < $5,000 $5,000 - $9,999 $10,000 - $14,999 $15,000 - $19,999 $20,000 - $24,999 $25,000 - $29,999 $0,000 - $4,999 $5,000 - $9,999 $40,000 - $44,999 $45,000 - $49,999 $50,000 - $54,999 26.0 21.15 1.78 1.65 1.04 1.2 7.65 5.28 1.90 1.76 0.44 0.22 0.40 1.8 0.55 0.45 $55,000 - $59,999 Annual Household Income $60,000 - $64,999 $65,000 - $69,999 $70,000 - $74,999 $75,000 - $79,999 $80,000 - $99,999 > = $100,000
Automobile Ownership Model NCSGRE March 2010 Page 5 while less than 5 percent of households with more than $45,000 income are without vehicles. Two-vehicle households are most commonly those with incomes of $20,000 to $0,000. Number of vehicles per household grows steadily with income, from 0.52 for households under $5,000 to 1.88 for households with $45,000 to $50,000 income to 2.66 for households over $100,000. The average for all households is 1.92 vehicles. Household vehicle ownership is directly related to the number of adults (for the purpose of this study, adults are defined as persons 16 years of age and older) in the household. Figure 2 shows that incidence of vehicle ownership and number of vehicles owned increases with number of adults. Of all households with one adult, 29.7 percent do not own vehicles, while only 8.76 percent of two-or-more-adult households do not own vehicles. The average number of vehicles owned increases with the number of adults in the household; 0.88 vehicles for one-adult households, 2 for two-adult household, 2.57 for three-adult households and about.29 for households with four adults or more. Figure 2 Percent of Households Owning One or More Vehicles by Number of Adults per Household 100% 0.00 0.06 0.45 4.42 0.00 1.97 0.07 0.0 0.05 0.49 0.66 0.00 1.9.5 0.52 0.58 0.00 4.77 1.17 0.00.79 7.06 11.40 1.27 16.62 1.27 14.29 14.12 80% 9 Percent of Households 60% 40% 20% 0% 21.18 0.90 5.0 46.02 56.77 5.88 56.5 11.76 20.12 26.19 21.18 2.01 1.41 29.7 16. 7.06 9.80 9. 1.27 5.86 8.76 11.76 6.41 0 1 2 4 5 8 7 6 5 4 2 1 0 Number of Adults (persons 16 years and older) As with household members having jobs, the number of household workers is closely related to vehicle ownership. Figure shows that both the percent of household owning vehicles and the number of vehicles owned are linked to the number of workers. Of the zero-worker households, 28.77 percent are without vehicles, while no households with four or more workers without a vehicle.
Automobile Ownership Model NCSGRE March 2010 Page 6 The relation between the average number of vehicles per household and the number of drivers is 1.5 for one-worker households, 2.14 for two-worker households, 2.81 for three-worker households,.89 for four-worker households, 4.50 for households with five or more workers. 9.89 percent of all households without any member with job own more than one motor vehicle in average; this can be explained by the fact that retired people are included in our sample. Figure Percent of Households Owning One or More Vehicles by Count of Household Members With Jobs 100% 90% 80% 0.00 0.07 0.57 0.14 0.00 0.07 0.11 4.54 2.65 0.16 0.05 1.74 0.48 2.55 0.2 0.00 0.00 1.69 0.84 4.18 4.68 10.80 9.28 18.5 16.17 27.99 18.14 9 Percent of Households 70% 60% 50% 40% 0% 20% 10% 0% 6.11 9.47 29.96 7.92 57.6 6.22 26.16 25.5 28.77 1.98 12.98 6.8 8.86 4.20 4.89 5.06 0.00 0 1 2 4 8 7 6 5 4 2 1 0 Count of Household Members With Jobs The number of licensed drivers in the household is closely related to vehicle ownership. Figure 4 shows that both the percent of households owning vehicles and the number of vehicles owned are linked to the number of drivers. Of one-driver households, 12.2 percent are without vehicles, while no households with three or more drivers are without vehicles. A somewhat surprising finding is that 2 percent of all households without any licensed drivers own at least one motor vehicle. Average number of vehicle per household closely follows the number of drivers, ranging from 1.05 for onedriver households, 2.1 for two-driver households,.00 for three-driver households, 4.00 for four-driver households, 4.86 for households with five or more drivers.
Automobile Ownership Model NCSGRE March 2010 Page 7 Figure 4 Percent of Households Owning One or More Vehicles by Number of Drivers 100% 80% 0.14 0.00 1.74 1.9 0.00 0.05 0.5 0.06 0.04 0.08 0.68 0.9 0.00.21 1.64 0.62 0.00 1.16 0.58 0.00.28 4.06 12.7 17.19 16.00 24.06 4.18 9 Percent of Households 60% 40% 20% 0% 50.6 7.97 98.12 72.98 64.91 22.7 22.7 26.8 26.05 0.00 12.2 11.89 11.6 4.06 1.56 2.05 0.00 1.74 0.00 0.00 0 1 2 4 5 8 7 6 5 4 2 1 0 Number of Drivers Figure 5 and Figure 6 presents data on vehicle ownership according to the FHWA household structure classification. The classification highlights some intuitive relationships. Single-adult households account for the greatest proportion of zero-vehicle households. Single-adult households where the individual is retired and without children are those most often without vehicles. Of these households 2.85 percent are without vehicles, compared to 15.0 percent for all households and 20.00 percent for single-adult households where the resident adult is in the work force. In single-adult households with children, the age of the youngest child bears a relationship to vehicle ownership. Of the single-adult households where the youngest child is under 6 years of age, 42.68 percent own no vehicle, compared to 2.47 percent of households where the youngest child is between 6 and 16 years, and 20.4 percent where the youngest child is 16 or over. The highest incidence of vehicle ownership occurs among multiple-adult households with children. Households that do not own a vehicle comprise only 6.21 percent of all multiple-adult households with the youngest child under 6,.72 percent of households with the youngest child between 6 and 16, and.87 percent of households with the youngest child 16 or over. Actual ownership rates in terms of vehicles per household range from 0.71 vehicles for singleadult with youngest child under 6 households to 2.99 for multiple-adult households with the youngest child 16 or older. The average vehicle ownership rate of single-adult households ranges from 0.71 where the adult s youngest
Automobile Ownership Model NCSGRE March 2010 Page 8 child under 6, to 1.4 where the youngest child is 16 or older. Ownership rates in multiple-adult households range from 1.72 where the household head is retired to 2.99 where the youngest child is 16 or older. Figure 5 Percent of Households Owning One or More Vehicles by Household Structure Figure 6 Percent of Households Owning One or More Vehicles by Household Structure Percent of Households 100% 80% 60% 40% 20% 0% 1.65 0.25 0.00 0.00 0.00 2.87 1.44 0.00 4.0 0.44 12.91 1.41 4.0 0.00.19 0.80 0.00 8.91 7.09 65.19 20.00 one adult, no children 4.90 42.68 one adult, youngest child 0-5 54.1 2.47 one adult, youngest child 6-15 Household Structure. 29.96 7.6 4.22 20.4 2.85 one adult, youngest child 16-21 one adult, retired, no children 9 8 7 6 5 4 2 1 0 Percent of Households 100% 90% 80% 70% 60% 50% 40% 0% 20% 10% 0% 1.90 0.6 0.15 0.19 0.10 0.00 0.6 0.12 0.00.52 1.86 0.18 0.00 0.49 0.25.80 4.91 2.88 2.58 1.66 0.00 8.82 9.2 10.29 19.65 18.9 16.24 25.57 5.50 57.12 48.8 5.24 2.80 14.5 12.87 8.46 7.8 5.89 6.21.72.87 2+ adults, no children 2+ adults, 2+ adults, 2+ adults, youngest youngest youngest child 0-5 child 6-15 child 16-21 Household Structure 45.8 0.15 9.19 2+ adults, retired, no children 9 8 7 6 5 4 2 1 0 Vehicle ownership increases with the level of educational attainment of the household head, principally because level of education is also tied to level of income. Both incidences of vehicle ownership and ownership rate increase with the level of education. As shown in Figure 7, 4.51 percent of households whose head did not finish high school are without vehicles. This proportion drops to 15.64 percent for those attending high school, and 2.5 percent for those have Bachelor s degree. Average number of vehicles owned is 1.00 for those households where the household head did not finish high school, 1.82 for those that attended high school, and 2.14 for those with Bachelor s degree.
Automobile Ownership Model NCSGRE March 2010 Page 9 Figure 7 Percent of Households Owning One or More Vehicles by Education of Household Head 100% 90% 80% 0.55 1.24 0.41 0.00 0.80 0.48 0.00 0.08 0.00 0.00 0.12 0.98 2.04 6.12 1.8 1.17 0.00 0. 0.66 0.00 0.00 0.20.1 2.29 0.78 0.72 0.00 7.46 5.28 7.16 4.48 8.61.65 8.17 12.9 18.0 16.40 17.97 15.98 19.75 16.45 16.7 70% 9 8 Percent of Households 60% 50% 40% 0% 20% 27.07 4.51 2.84 45.58 26.44 2.81 40.1 2.12 46.0 52.2 51.64 52.7 7 6 5 4 2 1 0 10% 0% 15.64 Less then High school high school graduate graduate 11.56 10.0 technical training college, no degree 19.26 5.06 Associate's degree 19.61 21.72 21.42 2.5 2.05 2.47 Bachelor's degree graduate school, no degree Graduate school degree Education of Household Head Figure 8 show the relationship between vehicle ownership and whether household is in urban or rural area. The results show that the households in rural area own more cars in average. The average vehicles per household is 2.5 in an urban cluster, 1.77 in an urban area, while in an area surrounded by Figure 8 Percent of Households Owning One or More Vehicles by Household in Urban/Rural Area Percent of Households 100% 90% 80% 70% 60% 50% 40% 0% 20% 0.00 1.14 0.08 0.0 0.04 0.44.98.52 1.28 0.00 5.26 0.7 2.97 0.84 4.46 10.80 14.44 1.16 15.04 19.74 2.95 27.02 42.10 5.2 25.42 51.2 40.67 9 8 7 6 5 4 2 10% 0% 14.20 1.70 12.66 In an Urban cluster In an urban area 10.5 7.80 0.00 0.46 In an area Not in urban area surrounded by urban areas 1 0 Household in Urban/Rural Area urban areas it is 2.51 and it is 2.8 in a non-urban area.
Automobile Ownership Model NCSGRE March 2010 Page 10 4 METHODOLOGY Multinomial logit has been used to model households decisions on vehicle ownership. We assume that the utility function is U in = V in + e in, and that each error component e in is distributed independently and identically in accordance with the extreme value distribution. Given this distribution for the unobserved components of utility, the probability hat the decision maker will choose alternative i is: P in ev i j J n e V jn for all i in J n (7) 5 MODELING FRAMEWORK 5.1 MODEL SPECIFICATION Having identified a set of relevant explanatory variables in Section, the next stage of model development is to identify a suitable modeling framework. This framework includes the identification of the relevant decision-maker, the alternatives available, and the functional form of the model. Although it can be argued that it is the individual (or the individual s employer in the case of company vehicles) who is responsible for specific ownership decisions, it is assumed that the household has overall responsibility for the number of vehicles owned. This assumption is not only reasonable but it is necessary given that the data does not identify the actual decision-maker.
Automobile Ownership Model NCSGRE March 2010 Page 11 The choice set for the household is well defined and includes zero, one, two, three and four or more vehicles. Consideration was given to extending the choice set to include five or more cars but examination of the data showed this to account for less than 2% of households. Although ownership levels are increasing, household size is also decreasing and therefore this sector of the market is not likely to be significant. Following the systematic assessment of each of the research issues, a final set of model was specified. The overall structure of the model assesses the household s decision to own zero, one, two, three or four or more vehicles by way of a Multinomial Logit Model. The final model specifications are shown below: Where, P i is the probability of owning each number of vehicles in the choice set (0, 1, 2,, 4+); P i depends on the factors that reflect the households need for vehicles and its willingness or ability to purchase vehicles. V i (the utility of ownership) denote the weighted sum of factors that affect households decisions. V 2 V0 0 V1 ASC1 a1 Inc b1 HS c1 Cld e1 Dri f1 Edu g1 Loc h1 HD m1 Rnt ASC2 a2 Inc b2 HS e2 Dri f2 Edu g2 Loc h2 HD m2 Rnt V ASC a Inc b HS d Wrk e Dri f Edu g Loc h HD m Rnt V ASC a Inc b HS d Wrk e Dri f Edu g Loc h HD m Rnt 4+ 4 4 4 4 4 4 4 4 4 Where: ASC i is alternative specific constant; Inc is the annual income of the household; HS is the household size; Cld is number of children in the household; Wrk is the number of workers in the household; Dri is the number of licensed drivers in the household; Edu is the education level of household head; Loc is the household location (on five level of variation ranging from urban to rural area); HD is the household density at block level; Rnt is the percent renter-occupied housing at block level; a, b, c, d, e, f, g, h, m are parameter vectors to be estimated.
Automobile Ownership Model NCSGRE March 2010 Page 12 5.2 MODEL ESTIMATION Table 1 presents the estimated model of vehicle ownership. The model is estimated on 20 observations and has very good level of fit with ρ 2 value with respect to constants of 0.4402. In terms of the estimated coefficients, most of them have the expected sign and values. For those which present significant t-statistics, the coefficients have intuitive meaning: (1) Coefficients of household income are positive and very significant. Meanwhile, the value of the coefficients is larger with respect to the households with more cars. Therefore households with higher income tend to own multiple cars, and the higher their income, the more likely that they will own more cars. (2) The variables relative to the number of household members have negative coefficient but result was not significant. () Number of children in a household is a significant factor entering onevehicle-alternative. (4) Households with more employees and drivers own more vehicles. The coefficients of number of drivers are extremely significant, which means that this attribute greatly influences the vehicle ownership in the household. Number of employees result to be significant in three-vehicle and four-or-more-vehicle alternatives only.
Automobile Ownership Model NCSGRE March 2010 Page 1 Table 1 Vehicle ownership model estimation Explanatory Variables Alternative Specific Constant Household Income Household Size Alternatives Estimated coefficient Standard Error t- statistic 1 Vehicle -1.04 0.647-2 2 Vehicle -7.66 0.758-9.7 Vehicle -12.02 0.909-1.2 4+ Vehicle -1.45 1.190-11. 1 Vehicle 0.175 0.070 4.7 2 Vehicle 0.071 0.040 7.6 Vehicle 0.854 0.046 8.8 4+ Vehicle 0.890 0.051 7.6 1 Vehicle -0.7756 0.156-5.0 2 Vehicle -0.206 0.126-1.6 Vehicle -0.2110 0.18-1.5 4+ Vehicle -0.292 0.167-1.8 Number of Children 1 Vehicle 0.511 0.157. Number of Employees Number of Drivers Education Level of Household Head Household Location Housing Density Vehicle 0.489 0.100.5 4+ Vehicle 0.4725 0.176 2.7 1 Vehicle.912 0.299 1.1 2 Vehicle 6.409 0.46 18.5 Vehicle 7.591 0.78 20.1 4+ Vehicle 8.0 0.417 19. 1 Vehicle 0.06871 0.0567 1.2 2 Vehicle 0.02271 0.0656 0. Vehicle -0.09585 0.0718-1. 4+ Vehicle -0.15570 0.0860-1.8 1 Vehicle 0.1887 0.149 1. 2 Vehicle 0.401 0.169 2.6 Vehicle 0.574 0.186.1 4+ Vehicle 0.5019 0.20 2.2 1 Vehicle 2 Vehicle Vehicle 4+ Vehicle - 0.0000609-0.00017750-0.0002050-0.00071770 0.000080-0.8 0.000092-1.9 0.000108 -.0 0.000175-4.1
Automobile Ownership Model NCSGRE March 2010 Page 14 1 Vehicle -0.02100 0.00478-4.4 Percent renter-occupied housing 2 Vehicle -0.02671 0.00556-4.8 Vehicle -0.0087 0.00644-4.8 Likelihood with Zero Coefficients = -4218.68 4+ Vehicle -0.04298 0.01010-4. Likelihood with Constants only = -64.957 Final value of Likelihood = -209.699 "Rho-Squared" w.r.t. Zero = 0.5165 "Rho-Squared" w.r.t. Constants = 0.4402 (5) When comes to the characteristics of the household head, the coefficients associated with household head s education level are significant. The coefficients in one-vehicle and two-vehicle alternatives are positive, as the higher the education level, the more likely that the household owns more cars. The coefficients in three-vehicle and four-or-more-vehicle alternatives are negative. All the education coefficients are not significant at the 95% level of significance. (6) We estimate three land use factors: location of the household (from urban to rural area), housing density and percent of rental properties. These variables have strong influence on household vehicle ownership. In particular, moving from urban to rural areas has a positive effect on the number of cars owned (as expected); housing density has a negative effect as well as the percent of rental properties. 5. MODEL VALIDATION For validation purpose (which is extremely important when the model is used to test policies), we re-estimate the model on 80% of the available observations in the dataset and we apply the model estimates on the hold out sample. The results show that the model does well in prediction. In Table 2 we report the actual choices, the choices predicted by the model and the difference between observed and predicted choices. It can be noted that we slightly under-predict the number of households with zero car and over-predict the number of the household with 1 or 2 vehicles. Differences are extremely small and less than 2.5%. We also apply the model by considering the dimension location of the household which is on 5 levels
Automobile Ownership Model NCSGRE March 2010 Page 15 of variation (urban, second city, suburban, town, rural); this is done because the low number of observations in rural areas might potentially compromise the ability of the model to correctly estimate the choices in remote areas. Table 2 Model Validation 0 Vehicle Household 1 Vehicle Household 2 Vehicle Household Vehicle Household 4+ Vehicle Household Cars per household Number of Household Number of Cars Actual Forecast Difference 28.04% 25.75% -2.29% 41.06% 42.75% 1.69% 24.2% 25.05% 0.7% 6.01% 5.67% -0.4% 0.57% 0.77% 0.20% 1.10 1.1 0.0 699 699 0 769 789.4 20.4 Again the differences in the number of cars observed and predicted for each area type are small; this attests the goodness of the model and the possibility to apply the model calibrated to test different policies. Table Model Validation (by land use categories) Urban Second City Suburban Town Rural Total Actual 0 Veh HH 2.0% 18.75% 1.64% 0.00% 0.00% 28.04% 1 Veh HH 42.54% 4.8% 6.07% 2.08% 0.00% 41.06% 2 Veh HH 20.85% 40.6% 44.26% 8.46% 66.67% 24.2% Veh HH 4.41% 6.25% 14.75% 0.77%.% 6.01% 4+ Veh HH 0.17% 0.00%.28% 7.69% 0.00% 0.57% Cars per household 0.98 1.4 1.82 2.2 2. 1.10 Number of Household 590 2 61 1 699 Number of Cars 579 4 111 29 7 769 Forecast
Automobile Ownership Model NCSGRE March 2010 Page 16 0 Veh HH 29.7% 12.81% 4.10% 0.77% 0.00% 25.75% 1 Veh HH 44.02% 47.19%.44% 26.92% 6.67% 42.75% 2 Veh HH 22.6% 2.50% 42.0% 8.46% 6.% 25.05% Veh HH.97% 6.56% 16.89% 24.62% 2.% 5.67% 4+ Veh HH 0.29% 0.94%.28% 9.2% 6.67% 0.77% Cars per household 1.02 1.6 1.82 2.15 2.0 1.1 Number of Household 590 2 61 1 699 Number of Cars 600.5 4.4 110.9 27.9 6.9 789.4 Difference 0 Veh HH -2.66% -5.94% 2.46% 0.77% 0.00% -2.29% 1 Veh HH 1.47% 12.81% -2.62%.85% 6.67% 1.69% 2 Veh HH 1.51% -8.1% -1.97% 0.00% -.% 0.7% Veh HH -0.44% 0.1% 2.1% -6.15% -10.00% -0.4% 4+ Veh HH 0.12% 0.94% 0.00% 1.54% 6.67% 0.20% Cars per household 0.04 0.01 0.00-0.08-0.0 0.0 Number of Household 0 0 0 0 0 0 Number of Cars 21.5 0.4-0.1-1.1-0.1 20.4 5.4 MODEL APPLICATION The model has then been applied to test a number of policies and to measure their effects on car ownership in the State of Maryland. We have tested the following scenarios: Change in housing density and in particular the effect of the increase in the actual values of density by 20%, 50%, 100%, 200%, 500%; Change in household income; we tested both a 25% decrease (reflecting a possible economic downturn) and a 25% increase; Change in land use; we assume that rural areas and towns become suburban area and that suburban areas become second city type; Change in land use: all areas become urban areas;
Automobile Ownership Model NCSGRE March 2010 Page 17 Unemployment: 10% of the household loose one worker and all the households loose one worker. Table 4 summarizes results obtained from scenario 1; it can be observed that small changes in housing density do not greatly affect car ownership and that we obtain a 4% reduction in the total number of cars in Maryland by doubling the actual density. This result can be explained by the fact that we are increasing density across the State and that more focused interventions could result in being more effective. In general we can observe that the number of household with zero or one car is increasing while household with multiple cars are decreasing in percentage. Table 4 - Scenario 1: Housing density Actual +20% +50% +100% +200% +500% 0 Veh Household 14.70% 14.47% 14.87% 15.55% 17.00% 21.79% 1 Veh Household.25%.9% 4.56% 5.55% 7.29% 9.86% 2 Veh Household 5.48% 5.28% 4.95% 4.2% 2.46% 27.45% Veh Household 1.1% 12.98% 12.49% 11.79% 10.69% 8.71% 4+ Veh Household.25%.4%.1% 2.88% 2.56% 2.20% Cars per household 1.57 1.57 1.54 1.51 1.45 1.0 Number of Household 20 20 20 20 20 20 Number of Cars 5218 5205.5 5128.1 5009.2 4798.1 404.7 Diff of Num of Cars 0-0.24% -1.72% -4.00% -8.05% -17.50% Income scenarios are presented in Table 5. As expected a decrease in household income will produce a decrease in the total number of cars owned by households in Maryland; a 25% decrease in household income is expected to lower the number of cars by about 4.5%.
Automobile Ownership Model NCSGRE March 2010 Page 18 Table 5 Income Factor Actual Income -25% Income +25% Value Difference Value Difference 0 Veh Household 14.70% 15.27% 0.58% 1.4% -1.6% 1 Veh Household.25% 6.04% 2.78% 1.40% -1.86% 2 Veh Household 5.48% 4.88% -0.60% 5.78% 0.0% Veh Household 1.1% 10.96% -2.5% 15.71% 2.40% 4+ Veh Household.25% 2.84% -0.41%.77% 0.51% Cars per household 1.57 1.50-0.07 1.65 0.08 Number of Household 20 20 0 20 0 Number of Cars 5218 4981.9-4.52% 548.2 5.08% An increase in household income of 25% will result into 5.1% more cars in our State. The most affected by this scenario are households with three cars. Similar analyses have been conducted by varying urbanization factors and unemployment rates; results relative to these cases are in Table 6. To facilitate the analysis of the results, we just report the total number of cars in the dataset, those predicted by the model under each of the scenario considered and the differences. When increasing urbanization the total number of cars and the number of cars per household decrease; an increase of the urbanization will generate a decrease in car ownership in suburban areas and towns, a relatively small increase is predicted for urban areas. These results are consistent for scenario and 4; in the latter case differences between actual and future conditions are very different resulting from the strong hypothesis that Maryland will become all urban. Rising rates of unemployment will produce a decrease of cars in suburban areas and town, but the scenario 10% rate of unemployment does not produce an overall decrease in the total number of cars. In order to quantify the effect on the population of Maryland in the last column of Table 6 we compare the actual number of cars in the State of Maryland and the predictions calculated by applying our model; it can be seen that even small effects predicted by our scenarios have strong effects on the total number of cars.
Automobile Ownership Model NCSGRE March 2010 Page 19 Table 6: Scenario -4-5 - Urbanization and unemployment effects Number of Cars Urban Second City Suburban Town Rural Total Sample Total Cars in State of Maryland(*) Urbanization - Scenario Forecast 1415.5 508.6 1841.8 1190.7 199.1 5155.,151,107 Actual 195 504 1887 121 201 5218,189,42 Difference 20.50 4.60-45.20-40.0-1.90-62.70-8,25 Urbanization - Scenario 4 Forecast 1415.5 494.1 1792.5 117.1 189.6 5029,07,908 Actual 195 504 1887 121 201 5218,189,42 Difference 20.50-9.90-94.50-9.90-11.40-189.00-115,524 10% Households loose one worker Forecast 1414.4 506. 1881.2 1210.9 206.6 5220,190,654 Actual 195 504 1887 121 201 5218,189,42 Difference 19.40 2.0-5.80-20.10 5.60 2.00 1,222 All Households loose one worker Forecast 198.5 496.2 187.6 1175.6 201.2 5108.9,122,746 Actual 195 504 1887 121 201 5218,189,42 Difference.50-7.80-49.40-55.40 0.20-109.10-66,686 *The population in State of Maryland is 5,296,486 and the average household size is 2.61 persons/household. Total number of households is 2,029,05. (from Census 2000) 6 CONCLUSIONS In this report we have presented a car ownership model for the State of Maryland. To our best knowledge this is the first model of this type developed for our State. The model has been calibrated on publicly available data (NHTS 2001) without the burden and the consequent cost to collect additional data. The sample has been sufficient to correctly estimate a number of relevant socio demographic and land use variables. The model has been tested on a hold out sample and differences between observed and predicted choices have turned out to be very small; this together with a number of statistical tests attests the reliability of the results obtained.
Automobile Ownership Model NCSGRE March 2010 Page 20 The model has then been applied, for demonstration purposes, to test a number of policy scenarios. We have mainly concentrated on housing density, income, urbanization and unemployment effects on car ownership. To summarize the results obtained we can say that housing density produces significant decrease in the number of cars owned if at least a 100% increase is applied. We suggest future testing focus on more localized changes where significant effects could be observed. Income has a strong effect on car ownership, while 10% random decrease of the number of workers in the household produces relatively small effects. The model can not only provide alone analysis, but also can be integrated into the statewide model or into MPO travel forecasting models. This can enhance the capability of modeling systems since car ownership influences trip making decisions