Is keeping one side of the escalator clear actually slower?
“Keep one side clear so people in a hurry can walk.” That has been the rule for decades. But judged on throughput alone, is it right?
Same number of people, same escalator. The only difference is how they ride it. Press start.
Which side clears everyone first?
Same number of people, same escalator. The only difference is whether one lane is kept clear for walkers.
What is happening
Keeping one side clear queues everyone into a single lane
When Hitachi Building Systems counted riders at Tokyo-area stations, 65% stood on the left and 35% walked on the right. Yet under the keep-one-side-clear convention, that 65% is funnelled into a single lane while the other lane sits half empty.
The queue grows in front of the standing lane only. In effect, riders throw away a lane they already have.
The walking lane is fast but sparse
Walkers leave two to four steps between them. Density is about half that of standing riders (0.27 vs 0.50 people per step). They move three times faster (45 m/min walking vs 15 m/min standing), so a single walking lane carries 0.54 × 3.0 = 1.62× what a standing lane does.
So the walking lane has high capacity but goes to waste when few people walk. That tension is the whole problem.
So the winner flips depending on how many people walk
Drag the slider to the right. Past roughly 50%, keeping one side clear wins. Push it further and the walking lane itself congests, so around 81% both-sides-standing wins again.
Keeping one side clear only wins inside that 50%–81% window. Japan’s measured 35% sits outside it.
Conclusion: standing on both sides is not universally right
“Keeping one side clear is inefficient” and “walking is obviously faster” are each half true.
The accurate statement is this: if more than half of riders walk, keeping one side clear wins. In Japan they do not. So it loses.
That leads somewhere interesting. The taller the escalator, the smaller the share of people who walk — about 44% on 23 m escalators versus 53–61% on 7–8 m ones. Not many people want to jog up two storeys. Fewer walkers means moving further from the 50% break-even, which means standing on both sides gets stronger.
Transport for London concluded that standing on both sides only helps on escalators taller than 18.5 m at peak times. Feed the height-dependent walker shares into this simulation and you get the same answer.
| Escalator | Share who walk | What this model says |
|---|---|---|
| 23 m (tall) | 44% | Both sides standing wins (by 12%) |
| 7–8 m (short, low end) | 53% | One side clear wins (by 6%) |
| 7–8 m (short, high end) | 61% | One side clear wins (by 28%) |
| Japanese stations | 35% | Both sides standing wins (by 30%) |
→ The precise claim is not “keeping one side clear is inefficient” but “it is inefficient on tall, crowded escalators.”
Facts and sources
Published measurements and figures this model computed are kept separate. The latter are not facts.
Published measurements (the model’s inputs)
- People per step: standing ≈10 per 20 steps (0.50/step); walking ≈5.4 per 20 steps (0.27/step) Source: Hitachi Building Systems (primary)
- Measured at Tokyo-area stations: 65% stand on the left, 35% walk on the right. Speeds: 15 m/min standing, 45 m/min walking. Moving 100 people took 104 s with one side clear vs 80 s with both sides standing (about 1.3×) Source: Mynavi News, 2024-05-20 (secondary; the original Hitachi material was not reached)
- Holborn station trial, Transport for London: 141 people/min standing on both sides vs 115 walking. The difference only appeared above 100 people/min, and the policy was never made permanent Source: Escalator etiquette (Wikipedia) / Londonist
- Height and walker share: about 53–61% at 7–8 m, about 44% at 23 m (2002 study, secondary)
Assumptions this model makes
Incline length 12 m (6 m rise at 30°), step depth 0.4 m, belt speed 0.5 m/s. These are typical values, not published figures from any source.
They do not affect the conclusion. The model turns on the ratio of density × speed, and both step depth and absolute speed cancel in that ratio. They only change the absolute times shown on screen.
What this model computed (not facts)
- One walking lane carries 1.62× what a standing lane does (density ratio 0.54 × speed ratio 3.0)
- At a 35% walker share, keeping one side clear takes 1.30× as long → this matches the published “about 1.3×”, which is what makes the model credible
- The winner flips at a 50% walker share, peaks at 61.8% (1.31× better), and flips back above 81%
⚠️ That 50%–81% window is not a published fact. It is what this model computes from the measured values above.
This simulation measures throughput only. The risk of falls, and consideration for people who can only use one hand, sit outside these numbers.