Urban access across the globe: an international comparison of different transport modes

The study investigates how accessibility to jobs varies with metropolitan population size across global cities, posing the question of whether larger cities enjoy increasing accessibility and how this relates to potential economies of agglomeration. It frames access (ease of reaching valued destinations) as a critical measure of urban efficiency produced jointly by land-use patterns and transport networks. Recognizing the growing importance of accessibility in transport, sustainability, and urban economics, the paper aims to quantify and compare population-weighted accessibility to jobs across cities worldwide and across transport modes, and to reveal how development patterns (density versus auto-oriented sprawl) influence accessibility outcomes. Using a universal, interpretable metric (30-minute cumulative job access), the study provides a basis to compare cities, assess the co-production of density and mobility in generating access, and understand implications for urban productivity and equity.

The paper situates its contribution within several strands of literature: (1) theoretical and empirical links between accessibility, land values, and urban structure, drawing on classical trade-offs between time and space and numerous hedonic studies evidencing positive correlations between access and land prices; (2) behavioral and economic outcomes affected by accessibility, including firm location choice, land development likelihood, commute mode choice, and transport emissions; (3) urban scaling and agglomeration research connecting city size to productivity, network structures, and transport system form; and (4) prior domestic and regional accessibility comparisons using cumulative opportunities in the US, Australia, Brazil, Canada, New Zealand, and Europe. The authors emphasize the advantages of cumulative-opportunities measures (absolute, comparable, interpretable) and note the gap in large-scale, multi-modal international comparisons. They also discuss regional differences in development patterns (e.g., US sprawl and road intensity vs. European compactness and denser networks) and their anticipated implications for accessibility.

• Measure: Cumulative job accessibility within a 30-minute travel time threshold for four modes (automobile, transit, walking, cycling), computed for 117 cities across 16 countries on six continents.
• Zone-based computation: For each origin zone i and mode m, accessibility equals the sum of jobs in destination zones j reachable within 30 minutes. A binary impedance function is used (1 if travel time ≤ 30 minutes; 0 otherwise). No distance decay is applied to maintain cross-city comparability given heterogeneous data and preferences.
• Aggregation: Zone-level access is aggregated to city-level averages. Where available, population-weighted averages are used to reflect the average experience of residents (working population weights for US, Australia, China, and Poland; total population weights for Brazil, Canada, Paris, and London). For African and Dutch cities, simple arithmetic averages are used due to data constraints (not population-weighted), likely yielding lower estimates than population-weighted measures for the same area.
• Temporal scope: Morning peak travel times. Automobile times reflect historical traffic and include recurrent congestion but exclude parking search. Transit times rely on GTFS (or equivalent) schedules including access, egress, waiting, and transfers, assuming schedule adherence. Walking and cycling use all legal links without congestion; walking speeds partly adjust for crossings but may overestimate actual walking access. Cycling uses all roads and may overestimate real access due to route choice under traffic stress.
• Scaling analysis: To quantify returns to scale between city population and access, cities are grouped by country and fit with a scaling model A = β0 P^{β1}, where β1 indicates returns to scale (β1 > 1 superlinear; β1 = 1 linear; β1 < 1 sublinear). Separate models are fit by mode and country. Goodness-of-fit (R²) and coefficients are reported.
• Modal comparison and ratios: The study compares access across modes and examines within-city disparities using ratios of automobile-to-transit and cycling-to-transit 30-minute access to assess how modal gaps relate to population and vary across regions.

• General scaling pattern: City-level access to jobs increases with population, but typically sublinearly (β1 < 1), indicating diminishing returns—doubling population yields less than double the access. The notable exception is transit in Chinese cities, which scales superlinearly (β1 ≈ 1.57), implying larger populations confer proportionally more transit-accessible jobs.
• Regional contrasts by mode:
  • Walking: Chinese and European cities exhibit markedly higher walking access for a given population; US cities have the lowest, reflecting lower densities and functional separation of land uses. Some outliers (e.g., New York, San Francisco-Oakland, Wellington, Sydney) align more closely with European patterns.
  • Cycling: Generally below auto but above transit; Chinese and European cities outperform similarly sized US cities. Oceania cities are comparable to the better US and Brazilian cases but below China/Europe. Cycling provides greater access than transit in every city with data; in Shanghai, cycling even exceeds automobiles where congestion depresses auto access.
  • Transit: Chinese and European cities have higher transit accessibility. Australian and Canadian cities are similar to one another and generally outperform most US cities of comparable size; Quebec City resembles US patterns. São Paulo and Rio de Janeiro underperform relative to their populations due to low densities in large territories and concentrated jobs, coupled with poor transport conditions and long commutes.
  • Automobile: Access increases with population everywhere. Chinese and European cities have the highest auto access at given population levels; US cities generally exceed Australia/Canada in auto access at the same scale, reflecting historical emphasis on automobility (e.g., Interstate system). New York has high auto access but is an outlier primarily for walking and transit.
• Scaling coefficients (examples from Table 1):
  • Europe: walking β1 ≈ 0.89 (R² ≈ 0.19); transit β1 ≈ 0.66 (R² ≈ 0.67); cycling β1 ≈ 0.56 (R² ≈ 0.63); automobile β1 ≈ 0.28 (R² ≈ 0.00).
  • United States: walking β1 ≈ 0.54 (R² ≈ 0.36); transit β1 ≈ 0.80 (R² ≈ 0.39); cycling β1 ≈ 0.54 (R² ≈ 0.49); automobile β1 ≈ 0.39 (R² ≈ 0.54).
  • China: transit β1 ≈ 1.57 (R² ≈ 0.40) superlinear; automobile β1 ≈ 0.65 (R² ≈ 0.82); cycling β1 ≈ 0.71 (R² ≈ 0.24).
  • Oceania: automobile β1 ≈ 0.52 (R² ≈ 0.84); transit β1 ≈ 0.47 (R² ≈ 0.75); walking β1 ≈ 0.36 (R² ≈ 0.56); cycling β1 ≈ 0.46 (R² ≈ 0.74).
  • Canada: automobile β1 ≈ 0.18 (R² ≈ 0.40); transit β1 ≈ 0.35 (R² ≈ 0.34).
  • Brazil: walking β1 ≈ 0.41 (R² ≈ 0.36); cycling β1 ≈ 0.45 (R² ≈ 0.37); transit β1 estimate inconclusive/negative due to very small sample (n=4).
• Correlation between population and access (across all cities): strongest for automobile (r ≈ 0.69), then cycling (≈ 0.55), walking (≈ 0.48), and transit (≈ 0.44).
• Modal disparity (using transit as benchmark):
  • Automobile vs. transit: Auto provides higher access than transit in all cities except Shanghai (auto ≈ 90% of transit at 30 minutes). Gap is largest in US cities; Australia/Canada show smaller but still substantial gaps. The auto/transit access ratio correlates weakly with population (r ≈ 0.10), indicating larger populations slightly narrow the gap.
  • Cycling vs. transit: Cycling exceeds transit in all examined cities. Ratios are relatively stable in Oceania and Europe (cycling about twice transit). Gaps are larger in US and Chinese cities. The cycling/transit ratio weakly declines with population (r ≈ 0.17), implying marginally smaller gaps in larger cities.
• Overall synthesis: Density and mobility co-produce higher access. US-style sprawl with intensive roads yields modest auto access for city size and low transit/walking access; Australia/Canada have lower auto access but better transit than US; compact Chinese/European cities with robust networks achieve the highest access for their population sizes.

The findings demonstrate that accessibility scales with metropolitan population but typically with diminishing returns, refining expectations about how agglomeration benefits materialize in access terms. The clear national and regional patterns indicate that urban form (density, land-use mix) and network provision (road and transit quality) jointly determine accessibility outcomes. Chinese and European cities show that compact development combined with intensive transport networks can yield high access across all modes; in contrast, US cities’ auto-oriented sprawl suppresses walking and transit access and only delivers modest auto access relative to city size. The mode disparity analyses suggest that larger cities can slightly reduce the auto–transit gap, but access equity concerns remain given transit’s lower reach relative to automobiles in almost all cases. These insights directly address the research question by showing that increasing population alone does not guarantee proportional accessibility gains; instead, planning choices that align density with high-quality multimodal networks are crucial to realize access benefits and the broader agglomeration advantages associated with urbanization.

This study provides the first large-scale, multi-modal international comparison of 30-minute job accessibility, benchmarking 117 cities across 16 countries for automobile, transit, walking, and cycling. It establishes robust scaling relationships between population and access, showing generally sublinear returns, with a notable superlinear case for transit in China. The work highlights distinct regional patterns, emphasizing that compact urban form and strong multimodal networks co-produce superior accessibility, while auto-oriented sprawl underperforms for transit and active modes. Policy implications include prioritizing coordinated land-use intensification and multimodal transport investments to enhance access and narrow modal disparities. Future research directions include: (1) linking accessibility with city-level indicators such as income, GDP per capita, transport emissions, and commute durations to quantify co-benefits and trade-offs; and (2) developing global time-series accessibility datasets to analyze the co-evolution of access with urban form and networks over time.

• City boundary definition (modifiable areal unit problem): Differences in delineation (e.g., exclusion of exurban areas) can bias access levels; while population-weighted measures mitigate this, cross-national inconsistency remains.
• Data consistency and coverage: Variability in census/job and population data accuracy and definitions across countries; informal economy jobs are generally excluded, potentially underestimating accessibility in some cities.
• Mode-specific measurement biases: Walking speeds and treatment of crossings may overestimate walking access; cycling uses all legal roads, overestimating practical cycling access under traffic stress; transit assumes perfect schedule adherence; automobile times exclude parking search.
• Temporal specificity: Estimates reflect morning peak conditions; accessibility varies by time of day and day of week.
• Small samples for some country–mode combinations (e.g., Brazil transit) limit inference precision for scaling coefficients.