The benefits and biases of seeing the world's cities through marathons

Abstract

Marathons are now common ways of seeing cities, yet little is known about how representative their routes are. Using 311 marathon routes across five continents, we compare landmarks and amenities along the course with those elsewhere in the same city, finding that museums are 15.7 times denser near the route and that the median city has about 8.5 times more luxury brands near the route than elsewhere in the city. These patterns persist under perturbed routes with the same start and finish lines: monuments and landmarks, in particular, are more prevalent on the race course than on similar alternative routes, suggesting that marathons function as intentionally selective urban portraits.

Figures (10)

• Figure 1: Marathon biases.A Considering the density of different monuments and attractions within 200 m of a marathon route, we see that routes are most skewed towards museums, reflecting centrality, but they are also denser with memorials and plaques, indicating marathons preference historic cores. B Using Paris as an example, we see that that marathon passes through an area with high concentrations of both attractions but also restaurants, although it then leaves these dense areas when it travels through the Bois de Vincennes and the Bois de Boulogne. C Adjusting for the skew towards density, we find that everyday services like schools and daycares are underrepresented but museums and galleries are overrepresented. D We permute each race while fixing start and finish lines and find that race courses are 0.9 and 2 standard deviations from our permutations at the median, for amenities and tourist attractions respectively; relative to these null models, Los Angeles is more biased towards attractions while Seville is more biased towards amenities.
• Figure 2: Luxury brands and marathon routes.A We create a list of luxury brands and observe density biases on-route and off-route, as above, finding the biases to be far higher than for other classes of amenity. B Considering the distribution of these biases across cities, we see that some marathons---like London's---have no luxury bias at all, while many Italian marathons---like Florence's---have a strong luxury bias. C Apple stores present an interest comparison because the brand makes a conscious effort to locate its stores as "destinations" rao2025toward, and we see that any given marathon route typically pass closer to an apple store than the average street in that city.
• Figure S1: Data description.A Locations of the marathons in our sample. The dataset is concentrated in Europe and North America but spans all inhabited continents. B Number of races by continent. Europe and North America account for most of the sample, with smaller samples from Asia, Africa, South America, and Oceania. C Distribution of route lengths derived from the GPX traces. Thankfully, most routes cluster tightly around the official marathon distance of 42.195 km, with a median of 42.4 km. Noise can be explained by imprecision in the GPX traces rather than true errors in the routes.
• Figure S2: Population and density.A Median on route to off route ratios for population density, building volume, building height, urbanisation, built surface, and vegetation (NDVI). Marathon routes tend to pass through denser, taller, and more urbanised parts of cities; vegetation is close to parity. Percent labels give the share of cities above parity. B and D City level comparisons for building height, vegetation, and population density. Most cities lie above identity for building height and population density, while vegetation is much more balanced.
• Figure S3: Parks and water.A Observed amenity ratios against the mean null ratio for each city. B The same comparison for landmark ratios. Axes are on log scales, and points above the dashed identity line indicate that the observed course is more biased than the average alternative route. This occurs in 68% of cities for amenities and 82% for landmarks. C Excess over the null by route type, measured as the observed ratio divided by the mean null ratio. Loop races tend to show more excess than point-to-point races, especially for landmarks; horizontal marks show medians and the dashed line marks parity. D Excess over the null against start finish gap. Routes with larger start finish gaps generally have less room for scenic detours, and landmark excess weakens as this gap grows.