Car Dependency Index (CDI) in Urban Mobility

• Car Dependency Index (CDI) is a quantitative metric that measures the necessity of private car travel in urban areas by comparing automobile and public transport accessibility or modal share.
• Methodologies for CDI computation integrate high-resolution spatial data, including hexagonal cell segmentation, travel-time networks, and POI mapping to capture diverse aspects of urban mobility.
• Empirical findings show that CDI is sensitive to income, spatial structure, and transit access, thereby providing actionable insights for benchmarking and policy planning in sustainable urban transport.

The Car Dependency Index (CDI) is a quantitative metric for assessing the relative necessity of private motorized travel, typically by car, within urban environments. CDI formulations quantify either the accessibility gap between automobile and public transport, the citywide fraction of trips made by car, or the share of population for whom cars are essential for daily mobility. The CDI bridges technical urban accessibility modelling, travel behaviour analytics, and applied policy evaluation, enabling rigorous benchmarking and equity analysis of urban mobility systems.

1. Formal Definitions and Methodological Variants

Multiple formulations of the Car Dependency Index appear in contemporary literature, each elucidating a distinct facet of car dependency.

Accessibility Gap Approach

In "Car Dependency in Urban Accessibility" (Campanelli et al., 1 Apr 2026), CDI is defined at fine spatial resolution as the normalized difference in opportunity scores between car and public transport modes for each city tessellation cell $h$:

$$\mathrm{CDI}_h = \frac{O_{h,\mathrm{Car}} - O_{h,\mathrm{PT}}}{O_{h,\mathrm{Car}} + O_{h,\mathrm{PT}}} \quad \in (-1,1)$$

where $O_{h,m}$ is the time-decayed number of Points of Interest (POIs) reachable from $h$ using mode $m$, formally:

$$O_{h,m} = \Big\langle \int_{t=0}^{\infty} f(2t)P_{h,m}(t_0, t)\,dt \Big\rangle_{t_0}$$

with

• $P_{h,m}(t_0, t)$: cumulative POIs within travel time $t$ departing at $t_0$
• $f(t)$: normalized time-decay kernel, typically $$\mathrm{CDI}_h = \frac{O_{h,\mathrm{Car}} - O_{h,\mathrm{PT}}}{O_{h,\mathrm{Car}} + O_{h,\mathrm{PT}}} \quad \in (-1,1)$$0 with $$\mathrm{CDI}_h = \frac{O_{h,\mathrm{Car}} - O_{h,\mathrm{PT}}}{O_{h,\mathrm{Car}} + O_{h,\mathrm{PT}}} \quad \in (-1,1)$$1 min
• $$\mathrm{CDI}_h = \frac{O_{h,\mathrm{Car}} - O_{h,\mathrm{PT}}}{O_{h,\mathrm{Car}} + O_{h,\mathrm{PT}}} \quad \in (-1,1)$$2: hourly samples from 8:00–22:00.

City-level CDI is computed as the population-weighted average: $$\mathrm{CDI}_h = \frac{O_{h,\mathrm{Car}} - O_{h,\mathrm{PT}}}{O_{h,\mathrm{Car}} + O_{h,\mathrm{PT}}} \quad \in (-1,1)$$3 where $$\mathrm{CDI}_h = \frac{O_{h,\mathrm{Car}} - O_{h,\mathrm{PT}}}{O_{h,\mathrm{Car}} + O_{h,\mathrm{PT}}} \quad \in (-1,1)$$4 is the population in cell $$\mathrm{CDI}_h = \frac{O_{h,\mathrm{Car}} - O_{h,\mathrm{PT}}}{O_{h,\mathrm{Car}} + O_{h,\mathrm{PT}}} \quad \in (-1,1)$$5.

Modal Share Formulation

In "Critical factors for mitigating car traffic in cities" (Verbavatz et al., 2019), the CDI is modeled as the citywide modal share of car drivers: $$\mathrm{CDI}_h = \frac{O_{h,\mathrm{Car}} - O_{h,\mathrm{PT}}}{O_{h,\mathrm{Car}} + O_{h,\mathrm{PT}}} \quad \in (-1,1)$$6 where $$\mathrm{CDI}_h = \frac{O_{h,\mathrm{Car}} - O_{h,\mathrm{PT}}}{O_{h,\mathrm{Car}} + O_{h,\mathrm{PT}}} \quad \in (-1,1)$$7 is the volume of car trips and $$\mathrm{CDI}_h = \frac{O_{h,\mathrm{Car}} - O_{h,\mathrm{PT}}}{O_{h,\mathrm{Car}} + O_{h,\mathrm{PT}}} \quad \in (-1,1)$$8 the total commuting population. This approach emphasizes the structural and behavioral determinants of car use, particularly access to mass rapid transit (MRT). In the limit where most residents with MRT access prefer transit over driving, and modal choice is shaped by station proximity $$\mathrm{CDI}_h = \frac{O_{h,\mathrm{Car}} - O_{h,\mathrm{PT}}}{O_{h,\mathrm{Car}} + O_{h,\mathrm{PT}}} \quad \in (-1,1)$$9, a closed-form approximation is

$O_{h,m}$0

where $O_{h,m}$1 is the fraction of population within $O_{h,m}$2 of an MRT station, and $O_{h,m}$3 is MRT station density.

Empirical Modal Split (ABC Fractions)

"Large cities are less efficient for sustainable transport: The ABC of mobility" (Prieto-Curiel et al., 2022) defines CDI (denoted $O_{h,m}$4) as the share of all weekly journeys in a city completed via private motorized vehicle: $O_{h,m}$5 where $O_{h,m}$6 is the active mobility share, $O_{h,m}$7 is the public transport share, and $O_{h,m}$8 (CDI) is the car modal share for city $O_{h,m}$9. Multimodal trips with any car segment are counted as $h$0.

2. Data Inputs and Computational Frameworks

Achieving accurate, scalable CDI estimation requires integration of high-resolution urban geospatial data, open transport networks, and flexible simulation procedures.

Accessibility-Gap CDI

• Spatial partition: Cities are divided into hexagonal cells ($h$1) at H3 resolution 9 (side ≃200 m).
• Population: Aggregated from WorldPop 100 m raster to hexagons.
• POIs: Extracted from OpenStreetMap, mapped from polygons to hex cell centroids.
• Travel-Time Networks:
  • For cars, travel times between all pairs are derived from OSRM routing on OSM, adjusted for city-specific congestion and a 15 min parking/walking buffer.
  • For public transport, pedestrian access times are calculated (up to 1.25 km), and scheduled transit routes constructed from GTFS using the Connection Scan Algorithm, with 15 min maximum transfer walks.
• Temporal resolution: Opportunity scores averaged over multiple departure times.

Modal Share and Survey-Based CDI

• Surveys: Citywide modal splits are derived from recent (pre-2020) travel surveys capturing all work journeys, aggregated into the ABC fractions.
• City boundary harmonization: Uniform use of administrative or functional urban area polygons to ensure comparability.
• Distance and station density: For theoretical models, population density and spatial distribution of MRT stations inform the population with PT access.

3. Key Properties, Interpretation, and Statistical Behavior

The CDI possesses distinct mathematical and urban-interpretive properties contingent on its definition.

Variant	Range	Interpretation
Accessibility-gap CDI	(–1, 1)	$h$2: car advantage; $h$3: PT advantage; $h$4: parity
Modal share (theoretical/emp.)	[0, 1]	Fraction of trip volume/commuters/car-dependent population

• Spatial granularity: The accessibility-gap approach produces a high-resolution surface ($h$5), enabling localized analysis of accessibility gaps within cities.
• Interpretability: Negative CDI values ($h$6) indicate superior public transport; positive values denote essential car reliance.
• Population weighting: Citywide CDI averages enable interurban benchmarking.
• Sensitivity to income: Empirical analyses reveal that CDI predicts car ownership after controlling for income.

Empirical findings across 18 case-study cities (Campanelli et al., 1 Apr 2026):

Rank	City (Boundary)	$h$7
1	Paris (Municipality)	–0.111
2	Zurich	–0.020
...	...	...
17	Chicago	0.270
18	Málaga	0.310
19	Rome	0.335

In the theoretical model (Verbavatz et al., 2019), modal share CDI robustly approximates $h$8 (fraction without rapid transit access), validated across 25 cities ($h$9).

Globally, modal split CDI ($m$0) averages 70% car share, with the US/Canada above 90%, and European cities typically spanning 50–80% (Prieto-Curiel et al., 2022).

4. Comparative Applications, “What-If” Scenarios, and Empirical Findings

CDI frameworks inform both descriptive benchmarking and prospective policy analysis:

• Spatial inequality: Central urban hexagons near high-capacity rail tend to exhibit $m$1, while peripheries display high car dependency.
• Equity: In Vienna, intra-district analysis confirms that higher CDI predicts higher car ownership, independent of wealth gradients (Campanelli et al., 1 Apr 2026).
• Inter-urban scaling: In non-US cities, CDI declines modestly with city population: a doubling yields a 6% reduction ($m$2). No such scaling exists in US cities; CDI remains high irrespective of population (Prieto-Curiel et al., 2022).
• Income elasticity: CDI (modal share) increases by 37% when per-capita income doubles (exponent $m$3) (Prieto-Curiel et al., 2022).

Rome Metro Expansion Scenario (Campanelli et al., 1 Apr 2026):

• New C- and D-line stops included in GTFS and travel-time graphs.
• Simulation yields citywide $m$4, with localized drops (within 5-min walking, $m$5).
• Top affected areas (≈11,000 residents) see CDI decrease from ≃0.333 to ≃0.110.
• Projected reduction: ≈60,000 fewer car commuting trips, corresponding to a 5 pp drop in car commuter rate for Rome.

5. Limitations and Analytical Caveats

Each CDI construction has structural and empirical constraints:

• Model assumptions: Theoretical models (e.g., (Verbavatz et al., 2019)) assume monocentricity, disregard buses/cycling, ignore adaptive mode switching, and treat transit access and population spatially independently. This may overestimate car-dependence in polycentric or multimodal cities.
• Data harmonization: ABC modal split CDI (Prieto-Curiel et al., 2022) is susceptible to survey comparability, inconsistent spatial delineations, and bias from multimodal journey assignment (any car leg counts as car).
• Urban structure: Accessibility-based CDI is sensitive to the accuracy of travel-time networks, POI data, and congestion proxies.
• Parameter uncertainty: Estimates such as station density ($m$6), value of time ($m$7), and network capacity ($m$8) introduce quantitative variability.

6. Policy Implications and Operational Use

CDI’s main operational value lies in its objectivity, comparability, and spatial granularity for urban mobility planning:

• Car-free zone viability: Hexagons with $m$9 can sustain car restrictions with minimal accessibility penalty (Campanelli et al., 1 Apr 2026).
• Targeting interventions: Overlaying CDI with socioeconomic data enables prioritization of transit expansions in high-dependency, low-income areas, thus promoting equity.
• Systemic vs. local investments: Isolated public transport upgrades yield localized CDI reductions; only large-scale, network-wide transit expansion shifts overall CDI significantly.
• Benchmarking and monitoring: The CDI enables cross-city comparison, temporal monitoring, and quantification of policy impacts on modal accessibility.
• Park-and-ride planning: Transition zones with moderate CDI values are candidates for multimodal interchange investments.

CDI provides a replicable, data-driven foundation for identifying where cars are a necessity rather than a convenience and for designing strategies to foster sustainable, equitable urban mobility systems (Campanelli et al., 1 Apr 2026, Verbavatz et al., 2019, Prieto-Curiel et al., 2022).