Should I need a reservation to use public transportation? Deciding who gets to use our public spaces, when.

“Nobody goes there anymore. It’s too crowded.” — Yogi Berra

People standing in line with face masks. Source: https://flickr.com/photos/29418416@N08/49463786343

Standing in long lines has been a new reality in our Covid world, whether for grocery stores, food banks, or voting. Is this our new reality for all shared spaces and resources for the social distancing future, from restaurants, gyms, and hiking trails, to public transportation and even our work places?

As a full shelter-in-place becomes politically and economically undesirable, and business-as-usual medically untenable, states and municipalities have started imagining what a partial re-opening looks like. The question of when to re-open what has dominated news coverage. Business owners and municipal government officials have been focused on the minutiae details of how to open safely (clean surfaces, mandatory masks, and the logistics of social distancing), if it is even possible.

One missing conversation is how we decide who gets priority to use our re-opened spaces, given that they’ll all be capacity constrained. Even with a robust testing and tracing infrastructure, we won’t be able to crowd into offices, restaurants, or subways like before — everything will operate at half (or less) capacity. Governments, companies, and other institutions must thus decide how to allocate what is now suddenly scarce: time-slots to use shared resources and public spaces. If we can’t all be somewhere at the same time, someone needs to decide who gets what, when.

This issue is not new: pre-Covid, we have waited in line at popular restaurants, been stuck in traffic in rush hours, or scrambled for reservations to use public spaces such as community banquet halls and basketball courts. Such limited capacity and resulting rationing will be the new reality for all our public spaces, as long as social distancing is necessary.

A strategy for potential over-crowding, of course, is either to do nothing and hope spaces don’t get too crowded, or to do almost nothing and have people wait in lines when they do get crowded. As I discuss below, this strategy has essentially worked for grocery stores, who now see minimal lines after the initial pains. However, it seems to have failed for many parks and beaches, which have re-opened with either few restrictions (Florida), with limited parking (Austin), or with “park ambassadors” who can close down a park if it is too busy (Seattle): over-crowding has led to officials re-considering their opening decisions, with some preferring that no one use these spaces over too many doing so.

Public transportation, office work-places, and many businesses won’t have the option of simply shutting down if they get too crowded, and simple line based solutions might be too inconvenient (do you really want to stand an hour in line to see if you can go into work, and then to enter your gym, and then to dine at a restaurant?).

So can we do better?

In my research, I design and build socio-technical systems to help groups of people coordinate. And in the past month I’ve talked with numerous local government staff members, store managers, and institutional leaders about their plans and experiences re-opening and managing crowds.

In this post, I’ll present some of the challenges and possibilities facing our leaders and us, as we try to navigate a newly capacity constrained world.

Allocating scarce resources is an old issue for economists (enough so that a classical definition of economics is “the study of … the allocation of scarce resources”). Researchers in my field, at the intersection of computer science and economics, have contributed to algorithms assigning medical students to residencies, kidneys to patients, affordable housing to those in need, and much more. From computer science, a real-time allocation mechanism balancing fairness and congestion underpins our modern internet.

While many of the ideas in these fields are too complex or burdensome to be used for everyday applications, they do provide lessons and prototypical solutions. Here, I describe three classes of solutions: (1) lines, (2) priorities and even/odd systems, and (3) reservation based systems. I discuss where each solution might be most applicable, and what new applications and systems they might enable.

Lines, information sharing, and selfish routing. The naive approach is to do (almost) nothing: if a given space (store, park, library, office, public transportation) is full, have people wait outside until it isn’t.

This approach is particularly powerful when some people can and are willing to shift their usage to off-peak times, and can learn patterns for when a place is busy. For example, people in my neighborhood learned that our Trader Joe’s has a long wait on weekend mornings, and so those who could shifted their shopping to weekdays and other non-traditional times — out of our own selfish desire to not wait in line, but collaterally to the benefit of others. This way, usage naturally balanced: according to a manager, last week the longest line was ~15 minutes, down from well over 40 minutes.

Some technical sophistication could further streamline this approach: real-time information could be shared about current wait times, or people could “virtually” wait online. Such solutions would allow us both to adjust our usage patterns faster, and to spend less time in physical lines.

Does this solution always work? Unfortunately, no. For example, in a pre-Covid world, we all both knew approximate rush hours in our respective cities and had instantaneous access to traffic via Google maps, and yet rush hours still occurred: many people would or could not shift their travel times. Our transportation systems were thus over-crowded, and traffic would back up. More recently, when it came to re-opened beaches and parks, nice weather and plenty of sunshine was too tempting to resist, and people were willing to brave crowds.

Priority passes, Senior Hours, and Even/Odd systems. If an unrestricted, line-based system does not reduce crowds even at a given space, an alternative is to apply the idea of grocery stores senior hours: someone with power declares that certain people receive priority access to a given space during special times. That access could come in many forms, from allowing people to skip lines to preventing everyone else from using the space at all.

Such systems already existed in our pre-Covid world. In many countries, including India and China, some cities adopted “even/odd systems”, in which only allowed half the population to drive each day, to limit congestion and pollution (those with odd license plates can drive on odd days, and vice versa). Many highway systems have “HOV” (high occupancy vehicle) lanes during rush hours, to encourage car-pooling. Disney World allows those able to pay to skip lines for popular rides.

Such systems can have substantial benefits over line-based ones: they can (equitably) reduce crowds, allowing for example everyone to use a park several times a week, without any crowding. They may further allow institutions to assert their legitimate priorities and support vulnerable populations, like senior hours do.

It’s easy to see how we can use these systems moving forward. When businesses open, we might need “medical and other essential worker cars” on public transportation, at the expense of workers can can more easily work from home. Workplaces, membership based gyms, and government spaces like parks and libraries might all need to adopt an “even/odd” system, in which each person is assigned the times or days in which they can use the service.

The key challenge of such systems, of course, is enforcement. Even/odd driving restrictions have had mixed success, for example, based on whether people can buy multiple cars or license plates. For public spaces with many entry and exit points, such as parks and beaches, restricting access might be physically impossible. If we have “essential worker” cars on public transportation, some people may just declare themselves essential to skip the regular line. From my conversations with city parks administrators, police may not be willing to enforce such micro-restrictions, and so we might have to rely on honest behavior.

The second, related challenge is how to set the priorities. The public will not comply with a system it deems to be illegitimate (for example, if NYC banned half their essential workers from using the subway each day, or otherwise prioritized white-collar office workers). Administrators in charge of public spaces need to be thinking through such priority systems today, and preparing both enforcement and to convince the public that they considered the trade-offs carefully and equitably.

Reservations and individual preferences. The above discussion misses a crucial aspect: people have preferences over when they use a space. An even/odd system that lets me to come in to work on Mondays and you on Tuesdays is terrible if you prefer Mondays and I prefer Tuesdays!

Fortunately, we have experience solving this problem: have people make reservations to use a space, with some way to deal with time-slots that are over-requested (first come first serve, a lottery, priority access, etc).

Such a system, properly implemented, would enable fine-grained control and sharing of our shared resources, and would combine the benefits of the two systems above: (1) like the naive solution of having people stand in line if a place is busy, people could decide for themselves when they want to use a space, and (2) like a priority or even/odd system, the administrator could control congestion and encode priorities for who deserves to use an over-demanded resource the most.

Unfortunately, this technical solution imposes substantial burdens on people, that need to be justified. For a public that is used to just running to the store or getting on the subway whenever convenient, having to sign-up ahead of time might seem unnecessary and over-engineered. There might be equity or privacy concerns, as not all of us have the technological access needed to consistently reserve time on a website or app. And, for important spaces, building and deploying something at the scale of a city is difficult on the necessary time-scale.

Where might we see such sign-ups? At the least, I expect industries used to reservations, like restaurants, to more widely adopt the practice. Gyms and other membership based services may also do so. Places where a signup is infeasible could adopt a hybrid even/odd that is respectful of preferences: each person simply is asked to use a service no more than a given number of times a week (without having to sign up ahead of time) — but this hybrid doesn’t work if everyone has the same preferences.

I’d like to see establishments experiment beyond a first come first serve reservation signup, to lottery or other systems. For example, a gym could run a lottery for who gets the coveted 6pm slot for a given day, with the daily losers receiving higher priority in future days. Grocery stores for which people are staying up until midnight to sign up for over-booked delivery time-slots could (and should) implement such a system today.

More interesting is whether offices will adopt a hybrid with priorities and even/odd assignments, with a reservation based system within teams that share a workspace. Consider a university’s research activities: in a phased re-opening, people conducting lab-based research requiring special equipment will receive priority. But those conducting theoretical research can’t stay home all the time. With highly heterogeneous preferences (dictated, for example, by one’s childcare situation and meeting schedule), universities will need to both allow us to choose when to come in, and somewhat restrict our ability to freely do so.

Such reservation systems could also be used to help re-purpose other spaces. For example, many banquet halls and streets currently lie empty. Restaurants would rush to use such spaces as overflow capacity if there was a simple matching mechanism to do so — a city could coordinate such usage fairly, using a reservation system.

As administrators of shared spaces make their re-opening plans, it will be tempting to hope that everything will be like grocery stores: that after some adjustment, a no-hassle, line-based solution will just work. I’m skeptical: I believe that even/odd or reservation based approaches will be worth the trouble in some settings.

Here are some general principles to consider when deciding on a system.

Priorities. Governments and institutions might be loath to seem like they’re prioritizing one group over another. However, every system prioritizes some people over others, whether implicitly or explicitly. Grocery stores without senior hours would be abandoning seniors who need to avoid crowded places, and long lines (such as to vote) prioritize those who have the time, energy, and ability to be in line — while such systems may seem like they’re working, that may only be because many have abandoned trying to use them.

Thinking through and communicating priorities now (such as getting us ready for “essential/medical worker cars” on public transportation) will save future trouble. After an institution formalizes its priorities, everything else is easy.

Enforcement, and off-loading capacity. There won’t be a one size fit all solution for our over-capacity shared spaces. Two features that will determine what type of system is best are enforcement capabilities, and whether peak demand can be “off-loaded” to other times (some people can shop at other times, but jogging in a park at midnight is hard).

As a rule of thumb, a naive or line-based solution will fail when a service is simply over-demanded at certain times, and even with long lines people won’t selfishly decide to some back at another time (or such non-busy times don’t exist). On the other hand, any non-naive solution needs enforcement, a clear communication of the system’s priorities, and potentially a new computational system to be built.

There are many details and design considerations for such sharing systems that I omitted here, and that need to be carefully considered — and I believe that the large economics and computer science literature on the subject can help.

If you are a policy-maker or administrator thinking through such challenges: I’d love to hear about what particular aspects you’re struggling with, and if there is a way I and my team can help. We’d be especially interested in building out a reservation or lottery system for a public space, or helping analyze data regarding daily usage patterns.

If you’re a researcher, in computer science, economics, or operations: there are many challenges and potential research questions here that need interdisciplinary attention. A partial list of questions:

  1. There is a large theoretical literature on non-monetary mechanism design (I ignored monetary solutions because for many of the applications above, they are dead on arrival, rightfully so). What from that literature can be used to design and build simple mechanisms usable by regular people?
  2. What level of heterogeneous preferences do you need for peak-demand offloading to work? What applications can we expect to meet that criteria? What can we measure or survey ahead of time to predict whether a line-based system work?
  3. How does an entity like a workplace or university combine information (priorities, office floor plans and spatial data, collaboration network data and meeting schedules, people’s preferences, capacity estimates, enforcement ability and communication costs) to decide who gets to come in on which day? More generally, are there usable allocation policies that jointly consider priorities, preferences, enforce-ability, and capacities over time?

I study CS/Econ and applications to socio-technical issues. Blog about books and technical issues. PhD Stanford, BS/BA UT Austin. gargnikhil.com

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store