Mode Share Validation

Mode Choice Model Validation

Published

January 15, 2025

Overview

This page presents the validation of modeled mode shares against observed data from the 2018 Utah Travel Survey and other sources. Mode share validation is critical to ensuring the model accurately represents travel behavior across different modes of transportation.

Validation Data Sources

  • Utah Travel Survey (2018): Household travel diary data collected from ~6,000 households
  • American Community Survey (ACS): Journey-to-work data by mode
  • UTA Ridership Data: Transit boarding counts by route and time period
  • Traffic Counts: Vehicle occupancy data from select locations

Regional Mode Share Comparison

All Trips

The following table compares modeled versus observed mode shares for all trip purposes:

Mode Observed (%) Modeled (%) Difference % Difference
Drive Alone 68.5 68.2 -0.3 -0.4%
Shared Ride 2 15.2 15.8 +0.6 +3.9%
Shared Ride 3+ 8.3 8.1 -0.2 -2.4%
Transit 2.1 2.0 -0.1 -4.8%
Walk 4.9 5.0 +0.1 +2.0%
Bike 1.0 0.9 -0.1 -10.0%
NoteValidation Criteria

The model meets FHWA validation guidelines, which recommend mode share differences of less than 5% for major modes. All modes are within acceptable ranges.

Home-Based Work Trips

Work trip mode shares are particularly important for peak-period forecasting:

Mode Observed (%) Modeled (%) Difference
Drive Alone 78.2 77.8 -0.4
Shared Ride 2 10.5 11.2 +0.7
Shared Ride 3+ 5.1 4.8 -0.3
Transit 3.8 3.7 -0.1
Walk 1.8 1.9 +0.1
Bike 0.6 0.6 0.0

Geographic Variation

Mode share validation by county demonstrates the model’s ability to capture spatial variation in travel behavior:

Transit Mode Share by County

#| eval: false

# Placeholder for actual visualization code
# import plotly.express as px
# import pandas as pd
#
# data = pd.DataFrame({
#     'County': ['Salt Lake', 'Davis', 'Weber', 'Utah'],
#     'Observed': [3.2, 1.8, 1.5, 0.8],
#     'Modeled': [3.1, 1.9, 1.4, 0.9]
# })
#
# fig = px.bar(data, x='County', y=['Observed', 'Modeled'],
#              barmode='group', title='Transit Mode Share by County')
# fig.show()

print("Chart placeholder - Transit mode share by county")
County Observed (%) Modeled (%) Difference
Salt Lake 3.2 3.1 -0.1
Davis 1.8 1.9 +0.1
Weber 1.5 1.4 -0.1
Utah 0.8 0.9 +0.1

Income Stratification

The model captures variation in mode choice by household income:

Transit Mode Share by Income Quartile

Income Quartile Observed (%) Modeled (%)
Low (< $35k) 4.2 4.0
Med-Low ($35-65k) 2.5 2.6
Med-High ($65-100k) 1.5 1.6
High (> $100k) 0.8 0.7
TipModel Sensitivity

The model appropriately shows higher transit usage among lower-income households and demonstrates reasonable sensitivity to household characteristics.

Time-of-Day Patterns

Mode share validation by time period:

Peak vs. Off-Peak Transit Share

Time Period Observed (%) Modeled (%)
AM Peak (6-9 AM) 3.5 3.4
Midday (9 AM-3 PM) 1.8 1.9
PM Peak (3-7 PM) 3.2 3.1
Night (7 PM-6 AM) 1.2 1.3

Walk and Bike Mode Shares

Active Transportation by Area Type

The model captures higher walk/bike shares in denser, more urban areas:

Area Type Walk (Obs) Walk (Mod) Bike (Obs) Bike (Mod)
Urban Core 8.5% 8.8% 2.1% 1.9%
Urban 5.2% 5.4% 1.2% 1.1%
Suburban 3.1% 3.2% 0.6% 0.5%
Rural 2.5% 2.4% 0.3% 0.3%

Journey to Work Comparison

Comparison with ACS journey-to-work data provides an additional validation point:

Commute Mode Share (Workers 16+)

Mode ACS 2019 (%) Modeled HBW (%) Difference
Drive Alone 75.8 77.8 +2.0
Carpool 13.2 16.0 +2.8
Transit 2.9 3.7 +0.8
Walk 2.1 1.9 -0.2
Bike 0.7 0.6 -0.1
Work from Home 5.3 N/A N/A
WarningMethodological Note

ACS data includes work-from-home respondents, while the model only estimates trips made. The comparison excludes work-from-home for consistency.

Validation Summary

Key Findings

Strengths: - Overall mode shares closely match observed data - Good representation of geographic variation - Appropriate sensitivity to household income - Time-of-day patterns well captured

Areas for Future Enhancement: - Bike mode share slightly underestimated in urban core - Continue monitoring as new bike infrastructure is added - Incorporate micromobility (e-scooters, e-bikes) in future versions

Statistical Metrics

Metric Value Guideline Status
Root Mean Square Error (RMSE) 0.38% < 1.0% ✓ Pass
Mean Absolute Error (MAE) 0.28% < 0.5% ✓ Pass
Maximum Difference 0.7% < 5.0% ✓ Pass

Technical Notes

Survey Expansion: Utah Travel Survey data was expanded to represent the full regional population using household and person expansion factors.

Model Outputs: Mode shares are calculated from the mode choice model outputs for all internal trips. External trips are excluded from this validation.

Comparison Period: Model base year is 2019; survey data is from 2018. Socioeconomic data was adjusted to 2019 levels for consistency.

Last Updated: January 15, 2025