Skip to content

Auto Ownership Model

Model Purpose

The Auto Ownership Model determines the number of vehicles available to each household, establishing a fundamental constraint that influences all subsequent travel decisions. This is typically the first choice model executed in the CT-RAMP system.

Model Overview

Purpose and Role

The Auto Ownership Model predicts household vehicle ownership levels based on:

  • Household characteristics: Size, income, workers, lifecycle stage
  • Accessibility: Transit and non-motorized accessibility to activities
  • Policy variables: Parking costs, vehicle taxes, fuel prices
  • Spatial context: Urban density, land use mix, walkability

Key Behavioral Assumptions

Rational Decision-Making : Households choose vehicle ownership levels that maximize their utility given their constraints and preferences

Long-term Decision : Auto ownership is treated as a medium to long-term household decision that provides the context for daily travel choices

Household-Level Choice : The decision is made at the household level considering all household members’ needs and constraints

Accessibility Trade-off : Households balance the cost and convenience of vehicle ownership against accessibility provided by other modes

Model Structure

SANDAG CT-RAMP Design Foundation

The Auto Ownership Model follows the original SANDAG CT-RAMP design specifications:

Model Formulation: Nested logit choice model with five discrete alternatives Decision Unit: Household level Timing: Two model instances - preliminary (pre-location choice) and final (post-location choice)

Choice Alternatives

The model includes exactly five ownership levels as specified in the original CT-RAMP design:

Alternative Description Typical Households
0 vehicles No household vehicles Urban, low-income, transit-accessible
1 vehicle One household vehicle Small households, mixed accessibility
2 vehicles Two household vehicles Multi-worker households, suburban
3 vehicles Three household vehicles Large households, low transit access
4+ vehicles Four or more vehicles Large, high-income, rural households

Two-Stage Implementation

The SANDAG design implements auto ownership in two stages:

Stage 1: Pre-Mandatory Auto Ownership (Model 2.1) - Used to select preliminary auto ownership level - Based on household demographics and general accessibility - Provides vehicle availability for workplace/school location choice - Excludes work/school-specific accessibilities

Stage 2: Final Auto Ownership (Model 3.2) - Re-run after work/school location choice - Incorporates actual work and school locations - Uses same variables as Stage 1 plus location-specific accessibilities - Alternative-specific constants may differ from Stage 1

Utility Function Structure

The utility of each alternative typically includes:

Household Demographics 

Utility = β₁ × HH_Size + β₂ × Workers + β₃ × Income + β₄ × Children + ...

Accessibility Measures 

+ β₅ × Transit_Accessibility + β₆ × Walk_Accessibility + ...

Spatial Variables 

+ β₇ × Population_Density + β₈ × Employment_Density + ...

Cost Variables

+ β₉ × Parking_Cost + β₁₀ × Vehicle_Cost + ...

Segmentation

The model often uses segmentation by:

  • Income categories: Low, medium, high income households
  • Lifecycle stage: Young adults, families with children, seniors
  • Geographic context: Urban core, suburban, rural areas

Technical Implementation

Model Type

Nested Logit Choice Model with five discrete alternatives and household-level choice

The original SANDAG CT-RAMP design specifies a nested logit formulation rather than multinomial logit, allowing for correlation among certain alternatives and more flexible substitution patterns.

Variable Categories

Based on the SANDAG design document, the model includes:

Demographic Variables - Household size (positive for higher ownership) - Number of workers (positive for higher ownership)
- Household income (positive for higher ownership) - Presence of children (mixed effects by age) - Presence of seniors (negative for higher ownership)

Accessibility Variables
- Transit accessibility logsum (negative for higher ownership) - Walk accessibility to retail (negative for higher ownership) - Auto accessibility advantage (positive for higher ownership)

Land Use Variables - Population density (negative for higher ownership) - Employment density (negative for higher ownership) - Retail density (negative for higher ownership) - Mixed land use measures (negative for higher ownership)

Cost and Policy Variables - Parking cost at home (negative for higher ownership) - Vehicle operating costs (negative for higher ownership) - Transit service quality (negative for higher ownership)

Sample Utility Function

V(0 vehicles) = β₀

V(1 vehicle) = β₁ + β₂×Workers + β₃×Income + β₄×HH_Size + 
                β₅×Transit_Logsum + β₆×Density + β₇×Parking_Cost

V(2 vehicles) = β₈ + β₉×Workers + β₁₀×Income + β₁₁×HH_Size + 
                 β₁₂×Transit_Logsum + β₁₃×Density + β₁₄×Parking_Cost

V(3+ vehicles) = β₁₅ + β₁₆×Workers + β₁₇×Income + β₁₈×HH_Size + 
                  β₁₉×Transit_Logsum + β₂₀×Density + β₂₁×Parking_Cost

Data Requirements

Input Data Sources

Household Data - Household ID and demographics - Number of workers and their characteristics - Household income category - Presence of children by age group - Presence of seniors

Person Data - Person age, gender, employment status - Student status and school location - Driver’s license availability

Zonal Data - Population and employment density - Land use mix and walkability measures - Parking cost and availability - Transit service frequency

Accessibility Data - Mode-specific accessibility measures - Auto, transit, walk, bike logsums - Level-of-service matrices

Required Preprocessing

  1. Accessibility Calculation: Compute mode-specific accessibility measures
  2. Density Calculation: Calculate population and employment densities
  3. Cost Data: Assemble parking and transportation costs
  4. Validation: Ensure data completeness and consistency

Model Outputs

Primary Outputs

Household Auto Ownership Level - Number of vehicles (0, 1, 2, 3, 4+) for each household - Choice probabilities for each alternative - Household-specific utility values

Derived Outputs

Accessibility Impacts - Auto ownership logsum for use in location choice models - Vehicle availability indicator for person-level models - Household mobility constraints

Validation Metrics

Aggregate Validation - Auto ownership distribution by income category - Ownership rates by household size and workers - Geographic variation in ownership rates

Disaggregate Validation - Individual household ownership prediction accuracy - Sensitivity to key policy variables - Comparison with observed survey data

Calibration and Validation

Parameter Estimation

Data Sources for Calibration - Regional household travel surveys - American Community Survey (commute mode data) - Consumer Expenditure Survey (vehicle ownership costs) - Local parking cost surveys

Estimation Approach - Maximum likelihood estimation using observed choices - Cross-validation with independent datasets - Sensitivity testing of key parameters

Validation Standards

Regional Benchmarks - Match observed ownership rates by geography - Reproduce income-ownership relationships
- Capture density effects on ownership

Policy Sensitivity - Reasonable response to parking cost changes - Appropriate sensitivity to transit improvements - Realistic response to demographic changes

Implementation Considerations

Configuration Parameters

Key parameters for regional adaptation:

Alternative Definitions - Maximum ownership level (3 vs 4+ vehicles) - Income category definitions - Geographic segmentation approach

Utility Function Specification - Variable transformation (linear vs log) - Interaction terms and segmentation - Constant term values for each alternative

Accessibility Integration - Logsum calculation methodology - Network representation and impedance functions - Time-of-day and mode specification

Performance Optimization

Computational Efficiency - Pre-calculate accessibility measures - Use vectorized utility calculations - Cache repeated calculations across households

Memory Management - Stream household processing for large populations - Efficient data structures for accessibility matrices - Garbage collection optimization

Common Issues and Troubleshooting

Convergence Problems

Symptoms: Model produces unrealistic ownership distributions Causes: Poor accessibility measures, incorrect parameter signs Solutions: Validate accessibility calculations, check utility function signs

Validation Failures

Symptoms: Model doesn’t match observed ownership patterns Causes: Incorrect segmentation, missing variables, poor calibration data Solutions: Refine segmentation scheme, add policy variables, improve calibration

Performance Issues

Symptoms: Slow model execution, memory problems Causes: Large accessibility matrices, inefficient calculations Solutions: Optimize accessibility calculation, use sparse matrices, parallel processing

Usage Examples

Basic Model Run

# Example model configuration
auto_ownership_config = {
    "alternatives": [0, 1, 2, 3, 4],
    "segmentation": "income_lifecycle",
    "accessibility_logsum": "mode_choice_logsum",
    "utility_specification": "full_specification.csv"
}

# Run auto ownership model
results = auto_ownership_model.run(
    households=household_data,
    accessibility=accessibility_data,
    config=auto_ownership_config
)

Scenario Analysis

# Test parking policy scenario
baseline_results = run_auto_ownership(base_parking_costs)
policy_results = run_auto_ownership(increased_parking_costs)

# Compare ownership distributions
ownership_change = policy_results.ownership_dist - baseline_results.ownership_dist
print(f"Policy reduces average ownership by {ownership_change.mean():.2f} vehicles")

Integration with Other Models

Downstream Dependencies

Models that use auto ownership results:

Coordinated Daily Activity Pattern (CDAP) - Vehicle availability affects activity pattern feasibility - Influences joint vs individual activity decisions

Tour Generation Models - Auto availability affects tour frequency - Enables longer-distance tours and activities

Mode Choice Models - Auto ownership level affects mode choice utilities - Determines auto availability for mode choice

Feedback Mechanisms

Accessibility Feedback Loop - Auto ownership affects overall accessibility - Changed accessibility influences future ownership decisions - Equilibrium achieved through iteration

Advanced Features

Policy Variables

The model can incorporate various policy levers:

Pricing Policies - Congestion pricing effects - Parking pricing policies
- Vehicle registration fees

Service Policies - Transit service improvements - Bike infrastructure investments - Car-sharing program availability

Behavioral Extensions

Household Coordination - Joint optimization of ownership and location - Intra-household vehicle allocation - Lifecycle-based ownership transitions

Uncertainty Modeling - Stochastic parameter variation - Confidence intervals on predictions - Sensitivity analysis automation

Model Foundation

Auto ownership is the foundation for all subsequent CT-RAMP models. Ensure this model is well-calibrated before proceeding to other components, as errors here propagate through the entire system.

Regional Adaptation

Auto ownership patterns vary significantly across regions. Local calibration using regional household travel survey data is essential for accurate modeling.

Related Components: - CDAP - Uses auto ownership results for household coordination - Tour Mode Choice - Auto availability affects mode choice utilities - Accessibility - Auto ownership affects accessibility calculations

Last updated: September 26, 2025