Urban highways are barriers to social ties

TL;DR

The study defines a Barrier Score $B_i = \frac{c_i^{\mathrm{null}}-c_i}{c_i}$ to quantify how urban highways constrain social ties, using a gravity-informed null model to compare observed cross-highway ties against a highway-agnostic baseline across 50 US metros. By overlaying geolocated social ties from Twitter (and Gowalla for robustness) with OpenStreetMap highways, the authors show a persistent, short-range barrier effect, strongest at distances under several kilometers and diminishing beyond ~20 km. Regression analyses at both city and census-tract levels indicate highways correlate with fewer social connections, with effects comparable to income and racial similarity, and the distance-interaction analyses reveal a nuanced pattern where highways hinder short-range ties but may aid longer-distance connections. The work further links high Barrier Scores to historical cases of racially motivated highway construction, arguing for evidence-based, reparative urban planning to address spatial social inequality, while acknowledging limitations in causality and data representativeness.

Abstract

Urban highways are common, especially in the US, making cities more car-centric. They promise the annihilation of distance but obstruct pedestrian mobility, thus playing a key role in limiting social interactions locally. Although this limiting role is widely acknowledged in urban studies, the quantitative relationship between urban highways and social ties is barely tested. Here we define a Barrier Score that relates massive, geolocated online social network data to highways in the 50 largest US cities. At the unprecedented granularity of individual social ties, we show that urban highways are associated with decreased social connectivity. This barrier effect is especially strong for short distances and consistent with historical cases of highways that were built to purposefully disrupt or isolate Black neighborhoods. By combining spatial infrastructure with social tie data, our method adds a new dimension to demographic studies of social segregation. Our study can inform reparative planning for an evidence-based reduction of spatial inequality, and more generally, support a better integration of the social fabric in urban planning.

Figures (15)

• Figure 1: The Highway Barrier Score measures the association between highways and social ties crossing them. Calculating the Barrier Score ${B_i}$ of a highway section $i$ follows four steps. The illustrated highway section consists of highway I-94 and the 8 Mile Road in Detroit. (A) Social ties: Count the number of times $c_i = 94$ that social ties (grey) between home locations of individuals (grey dots) cross the highway $i$ (red). (B) Rewiring: A spatial null model randomly rewires all social ties within a distance ring with a radius equal to the length of the original social tie. Within the ring, a random node is selected for rewiring with probability proportional to the local user population density, to reflect the spatial gravity law. The rewired null model ties remove any relationship between ties and highways because the rewiring is agnostic to highways. (C) Null model ties: Count the number of crosses $c_i^{\mathrm{null}} = 152$ of null model ties with the highway. (D) Highway Barrier Score: Calculate the Highway Barrier Score as ${B_i} = \frac{c_i^{\mathrm{null}}-c_i}{c_i}$. In this example, ${B_i} = +62\%$, which is the relative increase of social ties crossing the highway if ties were formed disregarding its presence. For illustration purposes, in this figure we only plot links that are fully within the view area.
• Figure 2: The Barrier Scores across the top 50 metropolitan areas in the US are consistently positive.(Left) Heatmap of all Barrier Scores $B(d)$ grouped into $0.5\,\mathrm{km}$ bins of social tie distance. Color denotes Barrier Score, square size denotes the fraction of social ties in each distance band relative to all ties in the city. All cities have positive Barrier Scores over most distances. Often, there is a smoothly reached peak distance, for example in Orlando at around $d_{\mathrm{peak}} \approx 1.5\,\mathrm{km}$. The top row labelled "ALL CITIES" reports the distance-binned Barrier Scores averaged over all cities. (Right) The bar plot labelled "CITY" reports the Barrier Score $B$ calculated considering all ties with distances up to $10\,\mathrm{km}$. All results shown are averaged over 15 randomized runs of the null model.
• Figure 3: Ordinary least squares regression across 50 cities reveals correlations between the Barrier Score and spatial features. The Barrier Score increases 1) with increasing highway length, 2) with decreasing fragmentation, 3) with decreasing user population density. The sketches on the right illustrate low and high values for the three features that are highway length, fragmentation, and user population density. Highways and user population are depicted via lines and dots, respectively. Grey backgrounds illustrate the signs of the regression coefficients. ***: p$<$0.01, **: p$<$0.05, *: p$<$0.1. Observations: 50. $R^2_{\mathrm{adj}}=0.231$.
• Figure 4: Historical case studies of highways associated with racial segregation. Highways are in color, following the color coding of Fig. \ref{['fig02:heatmap']} (red: positive Barrier Score, blue: negative Barrier Score, white: insufficient data). Brown rectangles denote historically relevant areas. Black dotted areas denote a city's districts with a black population share in the upper quartile. (A,B,C) Top 3 Barrier Scores: Cleveland, OH; Orlando, FL; Milwaukee, WI. Top Barrier Scores are consistent with these cities having well-known histories of highway-related racial segregation. (D,E,F) Interracial Barriers: Oklahoma City, OK; Cleveland, OH; Austin, TX. The barrier between Black and non-Black neighborhoods are clearly visible around I-235, the 8 Mile Road, and I-35, respectively. Detroit additionally features intraracial barriers around M-10, I-94, and I-75. (G,H,I) Intraracial barriers: Columbus, OH; Richmond, VA; Nashville, TN. Here the focus is on historically Black neighborhoods like Hanford Village, Jackson Ward, or Jefferson Street, respectively, that have been purposefully demolished via highway construction.
• Figure SI1: Social network size. Number of nodes and edges in the Twitter social network for the 50 largest metropolitan areas of the US.