By Christopher Cotton, Queen’s University

Yesterday, The New York Times published an article explaining why it is would be more efficient if the social norm involved everyone standing when riding escalators. The current norm in many countries involve those on the right standing, while leaving the left side of the escalators for walkers. The NYTs argues that we’d be better off as a society if both sides were used by standers alone.

The argument is based on an analysis conducted by the Capgemini consulting company last year, and published on their website in a blog post. The analysis was based on observed behavior in the Green Park Station in the London Underground, and a simulation based on this data conducted by the firm.

There does not appear to be anything inherently wrong with Capgemini’s simulation, or the assumptions based on their observations about escalator-rider behavior that they feed into their simulation. The issue is with the conclusions that they, the New York Times, and other outlets like Slate and Inside Science, draw from the analysis. Concluding that we shouldn’t walk on escalators is wrong. Why it is wrong becomes clear once one understand what’s going on behind the analysis.

The study

The analysis is based on a series of observations made while riding the long escalator in London’s Green Park Station. The team observed:

• People who stood on the right side of the escalator tend to leave a gap of 1 step between them.
• People who walked on the left side of the escalator tend to leave a gap of 2 steps between them.
• The journey up the escalator took standers 40 seconds, and the average walker 26 seconds.
• 40 percent of people display a preference for walking.

These observations were fed into a simulation of the Green Park Station escalator queue during Monday rush hour. They calculated that it took walkers, on average, 46 seconds to deal with the queue and escalator, and that it took standers, on average, 138 seconds to deal with the queue and escalator.

The analysis then considered how the wait time would change if both sides of the escalator were standing only. They show that the time dealing with the queue and escalator would be 59 seconds per person, on average.

So, it would take walkers 13 seconds longer, and it would take standers 79 seconds less. Massive gains for the 60% of the population that stand come at the expense of small losses to 40% of the population who walk.

This seems like clear evidence that the presence of walkers on escalators has huge negative externalities on the standers. So, let’s ban walking. Right?

Not so fast.

What’s driving the result?

The New York Times explains that walkers are inefficient because a walker keeps 2 steps between himself and the walker in front of him. A stander keeps only 1 step between himself and the stander in front of him.

This is not the real reason.

Although there is a bigger gap between the walkers, this is completely offset by the fact that walkers spend less time on the escalator. Walkers take up 50% more space on the escalator (every 3 stairs instead of every 2), but they also get off the escalator just over 50% more quickly than standers (26 seconds versus 40 seconds).

This means that the walking lane and the standing lane on an escalator are equally as efficient at getting people out of a queue and onto an escalator.

To illustrate this, imagine that an escalator ascents at 1 step per second. If you are in the standing line, you get on the escalator 2 second after the person in front of you. If you are in the walking line, you also get on the escalator 2 seconds after the person in front of you, on average. How is that possible if walkers need more space between them? It is because the walker in front of you takes a step up about once every 2 seconds. So, 2 seconds after he enters the escalator, he is on average 3 steps from the bottom. And, you step on.

If there are two queues of people at the bottom, one waiting to walk up the left side and the other waiting to walk up the right side, people will step onto the elevator from each queue at approximately the same rate. And those in the walking queue would reach the top more quickly, since they enter the escalator at the same rate but then ascend more quickly.

So, what’s driving the result that walking is inefficient? It’s the fact that less than half of the population is willing to walk up the escalator.

Because less than half the users are willing to walk, the standing lane tends to get more congested than the walking lane. There are periods when the walking lane is underutilized and the standing lane is backed up. This is where the issues come from.

The analysis concludes that society would be better off in a world in which everyone stood and no one walked on escalators. This would eliminate those periods of time in which the walking lane is underutilized and the standing lane is backed up. This would get masses of people up an escalator more quickly, on average.

An alternative conclusion

When only 40% are willing to walk, as the model assumes based on observations, the costs of congestion on those not willing to walk are huge. It takes standers more than a minute and a half longer to get up the escalator than it takes walkers (138 seconds versus 46 seconds).

This means that 60% of the population is willing to spend more than a minute and a half of their life standing in a queue in an underground tunnel to avoid a 26 second stair climb (the amount of time it takes walkers to ascend the escalator).

Improving the efficiency doesn’t require that the walkers stop walking. We can get slightly greater gains if we can somehow convince enough of the standers to stop standing and to start walking.

What we really need to do is convince enough of the population that they are willing to walk up the escalator when doing so saves time. Welfare could be most improved if in times when the walking lane is underutilized and the standing lane is backed up, enough people simply moved out of the standing lane and became walkers.

The time gains from this would even be larger than in a situation with no walking. This is because a walking gets one up the escalator more quickly than standing, and including a walking lane has the potential to shave time off of the average as long as it isn’t underutilized relative to the other lane.

To see this, think again about our hypothetical escalator that ascends at a rate of 1 step per minute, and imagine 202 people equally divided between two queues where (like at Green Park) each walker takes 26 seconds and each stander takes 40 seconds to reach the top once on the escalator. Suppose one of the queues is full of standers and the other full of walkers. As we describe above, each queue will step onto the escalator at a rate of one person every two seconds. Thus, the last person in each queue steps onto the escalator 200 seconds after the first person (100 people later times 2 seconds). But, the last person in the walking queue reaches the top 26 seconds after stepping on, while the last person in the standing queue reaches the top 40 seconds after stepping on.

This means that the last person in the walking queue reaches the top 14 seconds ahead of the last person in the other queue. And more importantly, so does every other person in the walker queue. The first person in the walker queue reaches the top 14 seconds ahead of the first person in the standing queue. The 17th person in the walker queue reaches the top 14 seconds ahead of the 17th person in the standing queue. And so on.

This 14 seconds per person time savings across the 101 people in the walking queue (1414 seconds = more than 23.5 minutes total) would be lost if everyone stood and the New York Times and the Capgemini propose.

This is what is wrong with the call to stop walking up escalators.

The analysis is right in that reserving half of the escalators is inefficient when too few people are willing to walk. But the best outcome doesn’t involve making everyone stand. The best outcome involves an intervention in the opposite direction. It involves encouraging more people who are otherwise standing to walk.

So, instead of convincing the 40% of the population who walks to stop walking, the better approach may be to convince some of the 60% of standers to not be as lazy, and to get 26 seconds of exercise.

(Having both sides of the escalator walk would shave off even more time, on average. But requiring both sides to walk is likely infeasible since some people are likely to find walking difficult or impossible given health, disability, stroller, or luggage. Although perhaps an everyone walks norm could be implemented in stations where those who cannot walk have the option of taking an elevator.)

Other issues

In economics, we worry about external validity of an experiment or analysis. This means whether our analysis of a specific situation can be used to make predictions about other situations. This is likely a significant concern for the analysis, which simulates the effects of the walking/standing social norm during rush hour on one of the busiest escalators in London. If we were to adopt a no walking norm, then even if there were gains during rush hour at the Green Park Station, there may be significant losses in time during different periods or at escalators where congestions is lower.

Finally, the analysis assumes that 40% of the population will walk and 60% will stand. A more detailed analysis should consider how these percentages change with the relative length of queue for both options.