This article is adapted from Shared Parking, Third Edition, published this spring by ULI, ICSC, and the National Parking Association. The book, 209 pages, is available in paperback ($155.95) or with Excel model ($649.95) at

Most experts expect that the part of the infrastructure first and most affected by autonomous vehicles (AVs) will be parking. By 2025, many new cars are expected to have “autonomous parking,” through which the car will be able to park itself in a parking facility without human intervention, even if the vehicle is not L5—the highest level of vehicle autonomy on a scale of L0 to L5.

Many manufacturers already offer some form of parking assistance, including Audi, BMW, Chrysler, Ford, Jaguar, Jeep, Tesla, Toyota, Volkswagen, and Volvo. In 2015, Mercedes-Benz became the first manufacturer to offer fully autonomous parking in a production model—i.e., the car can find a stall and park itself after dropping off the driver and passengers at the front door.

Unfortunately, most drivers think they can drive and park better than the systems on their cars or are frustrated by the much slower autonomous parallel parking operation; thus, a fair proportion of those who have this technology rarely use it. As noted in a 2017 article in Popular Mechanics (Ezra Dyer, “The Fallacy of the Self-Parking Car”), owners are far more likely to use the lane changing, backup cameras, speed control, or other advanced driver assistance system functions today. More recently, it is increasingly recognized that parking areas are highly complex, with pedestrians in the “roadway” and cars backing out of stalls. As with AVs in general, the timeline to widespread autonomous parking is still not predictable.

The first benefit of autonomous parking will be more cars parking in the same area. If passengers are dropped off and the car goes to park itself, no space is required for door opening at the stall. As many as six cars will be able park in the space of five. That alone is up to a 20 percent increase in parking capacity.

When planning new shared parking projects today, it is thus helpful to remember that parking capacity is likely to increase—fairly significantly—even as parking demand goes down. This is yet another reason to design a “just enough, no regrets” parking supply today. One would expect that perhaps as soon as 2025, building owners wishing to take advantage of the ability to gain spaces would start providing special areas for such vehicles to park and slowly expand those over time, reducing the parking area for driver-parked vehicles. At the same time, space for pickup and drop-off will be required, possibly in one location at grade or on each parking floor where autonomous parking occurs.

Obviously, it will take time for older cars without AV technology to be retired from service and time for the parking technology to be widely used. It will take even longer for a significant number of AVs to be driving around empty of passengers on public roadways.

If L4/L5 vehicles are sold to individuals, some families may be able to cut back to one car and use a ride-hailing service or other mobility options for some trips. At times, the family car will drive one family member to one activity and return to take another to a different activity. Parking can also be off site from destinations or the vehicle can go back home to park and perhaps recharge the battery during the workday.

However, the biggest game-changer for roads and parking demand (total cars parked at one time) is the marriage of ride-hailing services with AVs. Many expect that most urban dwellers will give up car ownership altogether while others are more skeptical that that vision will come to pass. Certainly many persons can benefit from improved mobility at lower cost. Aging baby boomers may be able to stay in their homes longer and get to doctor appointments and grocery stores more easily. Persons with mobility disabilities may have far better access to a mainstream lifestyle than with today’s paratransit services (a.k.a. dial-a-ride) at far lower cost to the public transit system. Teenagers can get to sports practices without disrupting parents’ work schedules. Lower-cost transportation options will provide improved mobility and choice of jobs for the working poor. Commuters can work or read and relax on the trip to and from the place of employment, which some worry will cause people to move yet farther out from central cities.

One other key point about the marriage of AVs and ride hailing exists: it increases vehicle-miles traveled (VMT) compared with the alternatives. There seems little doubt that congestion caused by transportation network companies (TNCS) such as Uber and Lyft is a growing problem in Manhattan as well as in certain other locations, including airport terminal curbs and in dining and entertainment centers. Studies of the potential for shared TNC rides almost uniformly show that even at maximum sharing, an overall increase in VMT will occur because of empty travel between paid trips. However, the fact remains that over two-thirds of U.S. VMT in 2018 was by private cars, and recovering excess space allocated to parking will significantly benefit the urban form.

Proponents of shared autonomous vehicle (SAV) rides believe that the best way to get benefit from AVs in cities is to have a significant migration to SAV rides rather than individual TNC rides. A 2018 Yale report (“Will Self-Driving Cars Usher in a Transportation Utopia or Dystopia?”) concluded that the “trifecta” of electric, shared, autonomous vehicles has potential for huge environmental benefits. And the reverse is true: if AVs are not shared and electric, the Yale report says, there will be “more gridlock, more pollution and more emissions . . . and to avoid the latter, public policy and regulations will have to force:

  • EVs [electric vehicles],
  • limited miles driven empty, and
  • incent SAV rides.”

Does the United States have the will to force SAV rides? Incentives (a.k.a. subsidies) for electric vehicles have not worked well. Although the industry has celebrated passing 1 million plug-in sales since 2010, less than 0.4 percent of the 260 million cars on the road as of the end of 2018 were plug-in. New York has been trying for 15 years but has yet to apply a congestion tax in Manhattan. The U.S. Congress cannot find the will to raise the gasoline tax to pay for crumbling bridges and roads.

When owning an electric vehicle is cost-beneficial (as many predict could occur within the next decade), the gas tax will significantly decline. This would occur at the same time city revenue from parking might decline, while at the same time significant need exists for upgrading and then maintaining the city’s infrastructure to be “smart” so that it can take advantage of all the potential benefits of AVs.

With the state of urbanization in the United States, it seems impossible for SAV rides to eliminate 90 percent of parking across an entire city, as projected by some studies and seized on by those promoting designing new parking facilities to be converted to other uses in the future. The academic studies on which this figure is based look at trips (not parking) that stay within a specific area, as noted by Wenwen Zhang and others in their 2015 journal article “Exploring the Impact of Shared Autonomous Vehicles on Urban Parking Demand” (Sustainable Cities and Society). First, this is not all trips, but only those that stay within the defined area. Second, the studies found that parking is reduced less than the trips or vehicles, by a factor of about 90 percent, as Zhang and Subrajit Gubathakurta found in their 2017 journal article “Parking Spaces in the Age of Autonomous Vehicles” (Transportation Research Record). If, indeed, 90 percent of trips in any one area are TNC, then the reduction in parking is about 80 percent.

“Transforming Personal Mobility,” a 2013 study of the potential vehicle ownership in the Ann Arbor, Michigan, metropolitan statistical area (MSA) found a much lower potential reduction in vehicle ownership, but only assumed individual TNC rides, not shared. That study found a potential reduction in cars on the road of 51 percent, with a reduction of 60 percent in privately owned vehicles. Using the 90 percent factor for parking, the reduction of parking across the MSA would be 54 percent. A 2018 study, “Driverless Future,” by Arcadis, HR&A Advisors, and Sam Schwartz Consulting, found that the reduction in parking will vary primarily according to residential density by area and overall. It found the following potential reductions in personal vehicle commuting trips in each of three MSAs: New York, 46 to 60 percent; Los Angeles, 36 to 44 percent; and Dallas, 21 to 31 percent.

And most important, will everyone even in dense urban areas choose shared rides? The proponents hope a significantly lower cost of SAV rides, compared with the cost of owning, operating, and parking a car, will cause the shift. A 2017 disruption scenario by the think tank RethinkX, offered in the study “Rethinking Transportation 2020–2030,” projects a 95 percent reduction in passenger miles traveled by 2030. A key assumption is that the cost of SAV rides will be 25 percent of that for owning a new car and half the cost of owning a paid-off vehicle. Further, RethinkX assumes that widespread TNC service occurs by 2021, resulting in the collapse of both new and used car sales by 2024 and the abandonment of existing vehicles.

Although the report concedes that rural residents will have little adoption of TNC rides, it did not seem to consider that 30 percent of VMT today is in rural areas. Further the RethinkX conclusion essentially requires not only the 53 percent suburban population, but also the 14 percent in urban clusters (towns of less than 50,000 population) to give up cars and use TNCs for all rides. Although rural population continues to slowly decline, 95 percent of passenger miles traveled by TNCs simply is not likely, particularly by 2030.

Assuming that people will give up cars and choose shared rides is, in itself, a huge leap. Studies of TNC use typically find two key motivators of TNC use: difficulty and/or cost of parking (which knocks out use for most local trips in suburbs) and avoiding drinking and driving. A 2017 survey of TNC riders at an airport—where parking costs for local residents are relatively high, as are alternative ground transportation costs for visitors to the region—found that 75 percent chose TNC rides for convenience and only 25 percent were motivated primarily by cost.

As Deloitte noted in the article “Tempering the Utopian Vision of the Mobility Revolution” in January 2019, “There are a few ‘immutable truths’ about consumer behavior: 1) Consumers are unwilling to compromise, 2) their usage patterns are difficult to change, and 3) they don’t like sharing.”
A 2018 study for the California Department of Transportation, “The Future of Autonomous Vehicles,” concluded, “Most notably, we find private ownership of AVs will prevail after a transition period.”

It is still useful for developers of parking to understand the possible magnitude of a future reduction in parking and a timeline. A reasonable consensus among business consulting firms seems to be that AV sales will be 15 to 20 percent of the market by 2030. Most, however, only discuss the percentage of AV sales and/or the percentage of vehicle sales to TNCs, not vehicles on the road. Walker Consultants used 2016 “high disruption” and “low disruption” projections of vehicle sales from McKinsey & Company (as seen in figure 1), population growth projections by the U.S. Census Bureau, and U.S. vehicle scrappage rates from automotive consultant IHS to project the vehicles on the road, which is rarely discussed in the literature. McKinsey estimated that for each AV sold to TNCs, sales of private vehicles would decline by 2.3 vehicles.

A point worth noting: the projected sales of L4/L5 AVs are 90 percent of the market in 2040 for the McKinsey high disruption scenario and only 10 percent for the low disruption scenario—an indication of all the hurdles that must be overcome for AV use to become common.

The vehicle sales calculations for the high scenario are represented in figure 2. The first thing gleaned from the sales figure is how many L0 to L3 vehicles will be sold in the United States (the gray area) before L4 and L5 vehicles are available, even with 15 percent of new vehicles being L4/L5 by 2030.

With an average age of over 11 years for privately owned vehicles and more than 20 percent still on the road at 20 years of age, it will take a long time to get non-AVs off the road, even at a high disruption scenario, as seen in the lower graph in figure 2. By 2030, 150 million L0 to L3 vehicles would be sold.

At full adoption, Walker’s calculations result in two-thirds of vehicles on the road being private and one-third TNCs. However, because of the miles driven per year by TNCs, 72 percent of miles would be accounted for by TNCs compared with 28 percent by private vehicles. Given that the final 2018 figures from the U.S. Department of Transportation indicate 30 percent of VMT was on rural roads, while accepting that rural population as a percentage of total population continues to decline, this is truly a reasonable high disruption scenario.

Walker then converted the vehicles on the road to high and low disruption scenarios for parking demand in the United States. The figure 3 projection for parking demand disruption without population growth would apply to the average U.S. building that has a fixed quantity of land use—that is, the average residential building or the average office building. The graph for parking demand with population growth would apply to places where activity grows with population, including downtowns, airports, universities, and the like.

The graphs represent an average reduction in parking across the United States; given that 53 percent of U.S. population is in the suburbs, it would tend to be applicable to many suburbs and the many cities without significant public (and particularly rail) transit in the United States. In such cases, parking demand would be projected to increase until about 2030 and then begin to decline, returning to roughly 2018 levels by 2050. Only then would the overall parking demand in the downtown of the average U.S. city begin to decline significantly.

Maximum Parking Reductions

Using the population by projected location of residency in 2018, one might conclude that the maximum reduction in vehicle ownership (and residential parking) and the maximum reduction in destination parking will be as follows:

  • Dense urban areas (7 percent of U.S. population): A high disruption scenario might indeed achieve a maximum of nearly 100 percent reduction in vehicle ownership and 90 percent reduction in destination parking by those residents who both live and work inside the center city limits (if SAVs are widely accepted to handle rides that are not convenient by transit). However, the parking by these residents is already reduced significantly by transit, walking, and biking.Figure 4 presents the car ownership in eight cities determined to qualify as dense urban areas. Significant variation still exists in the percentage carless, from 17 percent in Honolulu to 36 percent in Washington, D.C., and then a big jump to 55 percent in New York City.Further, some residents of these areas reverse commute to destinations outside the city limits. In addition, one must remember that less than one of five city trips are commute trips, according to the Federal Highway Administration.Take Boston as an example. According to the U.S. Census Bureau, it had an estimated population of 685,094 in 2017, with 4.84 million in the Greater Boston MSA. According to the 2017 American Community Survey, 34 percent of Boston (city) households are already carless. As of 2015, there were 380,000 off-street parking spaces citywide (not counting on-street spaces and those in private residential driveways and garages), with about 77,800 spaces in the downtown, according to the 2016 study Future of Parking in Boston by Better City and Nelson\Nygaard Consulting Associates Inc.About 37 percent of residents who work in the city commute by auto, compared with 64 percent for those commuting into the city. A 2017 study by the Boston Transportation Department, “Go Boston 2030,” notes that about 42 percent of the morning peak-hour trips to destinations in Boston are by residents of Boston. Conversely, about 98,000 city residents (who generate about 37 percent of the morning peak-hour trips originating inside the city) commute to work outside the city limits during the morning peak hour, largely by car. At least some of those people will not give up cars, so the reduction of residential parking within the city limits will be somewhat less than 90 percent. For the purposes of this discussion, a two-thirds reduction is assumed of vehicle trips by residents who reverse commute and by those who commute into Boston.Figure 5 summarizes a rough calculation of the reduction in morning peak-hour trips and a rough estimate of the potential reduction of parking demand for those vehicles in Boston. This is certainly not total parking but is a way of better understanding how the parking demand across even a dense urban city would decline somewhat less than 90 percent.For Boston, it is projected that parking at residences overall could decline 75 percent at a high disruption scenario, and parking for commuters and other trips that occur during the peak hour could decline 66 percent, which means parking for those users would be 34 percent of what it is today—significantly reduced, but not to 10 percent as estimated by many urbanists.A projection by the Boston Consulting Group in 2018 estimated that 30 percent of all trips within the city limits would ultimately be TNC, roughly half that of the peak-hour trip calculation in the figure.
  • Urban areas (13 percent of U.S. population): These could achieve a maximum of perhaps 67 percent reduction of vehicle ownership and perhaps up to 60 percent reduction in destination parking overall in a high disruption scenario.
  • Suburbs (53 percent of U.S. population): These could achieve an average reduction of 45 percent in car ownership (largely because families would be able to reduce the number of cars owned per household) and a 40 percent reduction in destination parking. A 2015 University of Michigan study, “Driverless Vehicles: Fewer Cars, More Miles”—which looked at the maximum reduction in U.S. vehicle ownership if one privately owned AV per household can make all trips—found a reduction of 2.1 vehicles per household today to 1.2 vehicles per household, a decline of 43 percent. Many of those in the study who could reduce car ownership by having a private AV handling all family trips are the same households that can reduce car ownership by using TNCs. In addition, note that paying for a more expensive AV is a lot easier if it eliminates the need for a second car as well.
  • Urban clusters and rural areas (27 percent of U.S. population): Virtually all analysts agree there will be little or no use of TNCs, much less SAV rides, in what the Census Bureau defines as urban clusters and rural areas. Parking may be reduced slightly if one privately owned AV will be able to drop the rider and go on another trip or go back home until needed. However, given the distances driven in such areas and the fact that parking is likely free, almost all private AVs are more likely to be parked at destinations in these areas.

Absorption of AV Impacts over Time

The timeline developed by Walker, based on multiple sources, shows maximum impact of AVs not occurring until 2050 even in the high disruption scenario. The parking market in a downtown or a campus with multiple parking facilities will absorb the changes in demand over time.

New developments will be built on surface lots or redeveloped sites and will use existing parking spaces that are underused because of the decline in parking demand. Parking may migrate to the perimeter, allowing a denser core.

It also must be remembered that significant TNC vehicle staging, cleaning, and recharging (if vehicles are electric) will be required, particularly if rush-hour commuting and events are to be accommodated. In addition, the increasing use of food delivery may require parking resources at both ends of the trip; these may displace vehicle trips to the same venue and be shorter in parking duration, but still reflect automobile parking needs.

Thus, a continuing need for some existing parking resources is likely, including some in prime locations, to facilitate shorter and faster “empty” trips even if parking demand for private vehicles declines.

It is certainly possible that parking structures can be converted to other uses in the future. However, given that most areas requiring parking structures have a mix of existing surface lots and parking structures of varying ages and conditions, it is far more likely that the older facilities will be redeveloped before it makes sense to entirely convert a parking structure built in 2020. It could be decades before a new parking structure built by, say, 2020 would need to be converted. (Figure 6 shows the parking supply in downtown Indianapolis as an example.)

Further, any future tenancy will be constrained by the initial parking configuration.


Consider the following urban living trends:

  • Urban living is holding its own but not growing as a percentage of total population.
  • Suburbs and less dense cities are growing, and rural population is declining.
  • Even in dense urban cities, most households are not yet carless.

One can hypothesize that as little as 20 percent of the U.S. population lives in truly dense urban neighborhoods where households could easily go carless because of SAVs, and then, likely—in fact, preferably—in combination with transit, walking, biking, and micro-mobility. Unless rides are shared, the cost of using TNCs will be about the same as that of owning a vehicle, according to most projections. Only shared rides would make TNC use cost-effective, and shared rides are simply not going to be practical for persons living in rural areas and urban clusters, nor for many living in suburbs.

Will all the remaining users in urban and dense urban areas choose shared rides? Not likely! That is not to say that ride hailing and autonomous vehicles will not have a significant effect. However, the idea that 90 percent of parking demand in a metropolitan area, much less the United States overall, will disappear is simply not achievable because of demographics and population density alone.

Shared parking will remain a key component of sustainable development for the future; however, “just enough” parking today is more important than ever before.

MARY S. SMITH, a certified Professional Engineer, is a senior vice president and director of parking consulting for Walker Consultants. She is a Fellow of the Parking Consultants Council and in 2018 received the International Parking Institute’s Lifetime Achievement Award. She also wrote the second edition of Shared Parking.