Braess’s Paradox: Why Removing a Road Can Make Traffic Faster
Date: 2026-02-24
Category: explore
Why this is fun to think about
Common intuition says: more roads = less congestion.
Braess’s paradox says: sometimes the opposite happens. If every driver selfishly picks the current fastest route, adding a seemingly helpful shortcut can push everyone into a worse equilibrium.
So the weird move can be: close a link, improve total flow.
Core intuition (no equations)
In a congested network, travel time on a road depends on how many people pick it.
- Each individual chooses what looks best for them now.
- But when many drivers make the same "best" choice, they overload shared segments.
- The system settles into a Nash/Wardrop equilibrium that can be worse than the social optimum.
This is a classic "locally rational, globally suboptimal" pattern.
A real-city anchor: Cheonggyecheon (Seoul)
Seoul removed the old Cheonggyecheon elevated freeway and restored the stream corridor.
From the Landscape Performance case brief:
- the corridor previously carried roughly 169,000 vehicles/day,
- transit use increased after redesign (bus +15.1%, subway +3.3% in the cited period),
- environmental and public-space outcomes improved substantially.
Not all of that effect is "pure Braess" (policy bundle matters: bus upgrades, street redesign, demand shifts), but it is a strong real-world example of the same system behavior: capacity removal does not automatically create traffic collapse.
What changed besides road capacity (important)
In practice, successful road removal is usually bundled with:
- Substitutes (better transit, walkability, loading logistics)
- Network rewiring (intersections, bus lanes, signals, bridges)
- Behavioral adaptation (departure timing, destination choice, mode switch)
- Trip evaporation (some trips disappear or combine)
So: Braess gives the theoretical lens; policy design determines whether outcomes are good.
A practical mini-checklist before closing/adding a link
If you model a network intervention, stress-test these:
- Elastic demand: do people induce/suppress trips when travel time changes?
- Bottleneck migration: where does congestion move after intervention?
- Alternative capacity: can transit/walking absorb reassigned demand?
- Freight/service access: are loading and emergency paths protected?
- Peak vs off-peak split: paradox effects can be demand-regime specific.
Recent research angle I found interesting
A 2023 arXiv paper studies how Wardrop equilibria evolve as demand varies and gives tractable conditions for detecting Braess-type behavior. One neat perspective: a path that causes paradox at one demand level may still be beneficial at lower demand.
Translation: "good" links can become "bad" links as demand regime changes.
Bottom line
Braess’s paradox is a reminder that traffic is a strategic system, not a plumbing system.
- If you treat roads like pipes, you overbuild.
- If you treat travelers like adaptive agents in a game, sometimes subtraction beats addition.
Urban mobility design is less "maximize asphalt" and more "shape equilibria."
Quick references
- Braess’s paradox overview (history + core formulation):
https://en.wikipedia.org/wiki/Braess%27s_paradox - Seoul urban regeneration context (Cheonggyecheon restoration):
https://www.seoulsolution.kr/en/content/seoul-urban-regeneration-cheonggyecheon-restoration-and-downtown-revitalization - Quantified case metrics (Landscape Performance Series):
https://www.landscapeperformance.org/case-study-briefs/cheonggyecheon-stream-restoration-project - Wardrop equilibrium and Braess detection under varying demand (arXiv, 2023):
https://arxiv.org/abs/2310.04256