Parking Minimum Elimination: A Pilot Evidence Review
Parking Minimum Elimination: A Pilot Evidence Review
What the two-axis framework reveals about a policy that is clearer than it looks, and murkier than it looks, at the same time
Suzanne Childress | Unicairn | April 2026
Parking minimum elimination is one of the most popular zoning reforms in the country right now. Buffalo did it citywide in 2017, followed by Minneapolis, San Jose, Austin, and dozens of others. The policy has unusual bipartisan appeal: urbanists want less driving, libertarians want less regulation, and housing advocates want cheaper construction. Everyone has a reason to like it.
The evidence base is surprisingly uneven. The policy’s effects on parking supply and housing costs are well-documented. Its effects on driving are not. And the gap between those two findings tells you something important about how transportation evidence works and why we need better frameworks for reading it.
This is the second pilot application of a two-axis evidence framework I am developing, which assesses interventions on both parametric confidence (how strong is the evidence?) and structural uncertainty (does the evidence answer the question practitioners actually face?). The first pilot reviewed Safe Routes to School. This one is a different animal.
What the policy is
Minimum parking requirements are local zoning regulations that require developers to provide a specified number of off-street parking spaces with every new building, typically calibrated to the building’s square footage or number of units. Most American cities adopted them in the mid-twentieth century, often based on the Institute of Transportation Engineers’ Trip Generation Manual, which was calibrated for low-density suburban development with ample free parking and no transit. Critics, most prominently Donald Shoup, have argued for decades that these requirements force developers to overbuild parking, increase housing costs, subsidize driving, and reinforce car dependence.
Parking minimum elimination removes the mandate. Developers can still build as much parking as the market demands. They are no longer required to build more than that.
Link 1: Does eliminating minimums reduce parking supply?
Yes. This is the strongest link in the causal chain, and the evidence is close to unambiguous.
The best study is Gabbe, Pierce, and Clowers (2020), which analyzed 868 multifamily developments comprising 60,361 housing units permitted in Seattle between 2012 and 2017, after the city reduced or eliminated parking minimums in large parts of the city. Parking minimums were the single largest predictor of how much parking developers actually built. After the reform, developers built 40% less parking than would have been required under the old rules, resulting in roughly 18,000 fewer parking spaces at an estimated savings of $537 million in construction costs. A third of buildings provided exactly the number of spaces the new minimums required, and many others provided more than the new requirement but less than the old one, suggesting that the old minimums were binding constraints.
In Buffalo, 14 mixed-use projects studied after the 2017 reform provided 53% less parking than previously required, with four projects building no parking at all.
Parametric confidence: High. Multiple cities, consistent direction, large sample in Seattle, clear mechanism. The finding that minimums are binding constraints on developer behavior is well-established.
Structural uncertainty: Low. The causal model is straightforward: remove the mandate, developers build less parking. No system-level feedback is likely to undermine this finding.
Link 2: Does less parking reduce car ownership and driving?
This is where the evidence gets complicated, and where the two-axis framework earns its keep.
The most cited study on parking and driving is McCahill, Garrick, Atkinson-Palombo, and Polinski (2016), which analyzed parking supply, automobile mode share, and Census commute data across nine mid-size American cities at three time points from the 1950s through the early 2000s. They found that an increase in parking provision from 0.1 to 0.5 spaces per person was associated with a roughly 30 percentage-point increase in automobile mode share, and used Bradford Hill causality criteria to argue that the relationship is likely causal.
This is a valuable finding, but it answers a different question than the one practitioners face. The McCahill study shows that building more parking over decades was associated with more driving. The policy question is whether removing parking requirements will reduce driving. These are not the same question, and the difference matters.
The relationship between parking supply and driving may be asymmetric. Over the second half of the twentieth century, abundant free parking enabled car-dependent lifestyles, sprawling land use patterns, and household location decisions organized around automobile access. Removing a parking requirement does not reverse those decisions on the same timescale. A household that already owns two cars, lives in a suburb, and has organized daily life around driving does not sell a car because a new building down the street has fewer parking spaces. The built environment and the behavioral patterns it supports were constructed together over decades, and the evidence that they can be deconstructed together by reversing one input is largely absent.
A Los Angeles County study using the 2017 National Household Travel Survey found that constrained on-site residential parking (less than one space per dwelling unit) was associated with a 10-23 percentage-point decrease in VMT. But this is cross-sectional: it compares households that already live in constrained-parking buildings with households that don’t. Self-selection is a major concern. People who want to drive less may choose buildings with less parking, and the study cannot fully separate the effect of the parking constraint from the preferences of the people who selected into it.
Parametric confidence: Low-moderate. The association between parking and driving is well-documented. The causal direction is plausible and supported by longitudinal evidence in the expansion direction. But no study has demonstrated the reverse: that reducing parking supply causes less driving.
Structural uncertainty: Moderate-high. The dominant concern is asymmetric reversibility. The evidence shows that building parking over decades contributed to increased driving. It does not follow that removing parking requirements will produce the reverse effect on a policy-relevant timescale, because the intervening variables (land use patterns, household location decisions, car ownership) change slowly and may not respond symmetrically.
The housing evidence is stronger
Parking minimum elimination has clearer evidence as a housing policy than as a transportation policy.
Construction cost savings are real and well-documented: the Seattle reform saved an estimated $537 million across 18,000 fewer spaces, roughly $20,000+ per unit in avoided parking construction. A San Francisco hedonic study found that off-street parking adds $39,000-$46,000 (12-13%) to home prices. Based on typical affordable housing development costs, one required parking space per unit increases total costs by approximately 12.5%, and two spaces can increase costs by up to 25%.
In Minneapolis, after parking reforms, developers proposed projects with fewer spaces, and new studio apartments that had typically rented for $1,200 per month were offered for less than $1,000. A Denver analysis suggested that eliminating minimums could increase housing production by about 12.5%. A Swedish study estimated that parking minimums reduce housing stock by 1.2% and increase rents by 2.4%.
The evidence is consistent and the mechanism is direct: parking is expensive to build, the cost is bundled into rents, and removing the mandate lets developers and the market decide how much parking to provide. That said, no rigorous difference-in-differences study has compared rents in a city that eliminated minimums against a matched control city. Most of the housing cost evidence is either hedonic (measuring the price premium of parking in cross-section), theoretical (construction cost per space multiplied by required spaces), or anecdotal (individual developments in specific cities).
Parametric confidence: Moderate. Consistent findings across multiple cities and methods. Direct mechanism. But lacking a rigorous causal evaluation of rents after a policy change.
Structural uncertainty: Low-moderate. The causal model (remove mandate → less parking built → lower construction cost → lower rents) is direct and does not depend on system-level feedbacks.
The bigger picture: we built this, and we don’t know how to unbuild it
The United States devoted the second half of the twentieth century to constructing a built environment organized around automobile access. The results are measurable. The U.S. has roughly 300 square meters of impervious paved surface per capita, more than double the Netherlands and triple Japan. Roads represent about 10% of land area in fully developed areas, and parking can consume up to 50% of land in commercial centers. Roughly 65% of America’s total impervious cover consists of streets, parking lots, and driveways. Donald Shoup estimated that the total land devoted to parking in the U.S. was larger than the state of Connecticut, and that was before the last two decades of growth.
The human costs of this system are not speculative. Pedestrian deaths in the United States increased 77% between 2010 and 2023, while all other traffic fatalities rose 22%. The U.S. is now a global outlier: pedestrian death rates have been climbing here while declining in virtually every other high-income country. Nearly two-thirds of pedestrian deaths in 2023 occurred in locations without a sidewalk. Fatal pedestrian crashes at night rose 84% over the same period, driven in part by the shift toward larger vehicles (SUVs and pickups now account for 54% of pedestrian fatalities). The people most exposed to this system’s failures are the people with the fewest alternatives: low-income households, children, the elderly, people without access to a car.
We know this system is dysfunctional. We know it produces pollution, sprawl, inequitable access, and a level of traffic violence that no other wealthy country tolerates. The question is whether we know how to fix it, and the honest answer, based on the evidence reviewed here and in the companion SRTS review, is that we are far less certain than we should be.
Asymmetric reversibility: a pattern worth naming
The parking minimum case illustrates what may be a general pattern across transportation interventions: asymmetric reversibility. The evidence that adding an input (parking, road capacity, highway access) caused an outcome (more driving, sprawl, car dependence) over decades does not mean that removing the input will reverse the outcome on a policy-relevant timescale. The built environment and the behavioral patterns it supports were constructed together. They do not necessarily deconstruct together.
Induced demand from highway capacity expansion is the canonical example. The evidence that adding lanes increases VMT is strong. But the reverse has not been demonstrated in the same way. Road diets (converting four-lane roads to three-lane configurations with bike lanes or wider sidewalks) have strong evidence on safety: FHWA documents 19-47% crash reductions, and individual projects show dramatic results (65% crash reduction on Ocean Park Blvd in Santa Monica, for example). What road diets do not have is evidence on VMT reduction. The studies measure crashes, speeds, and corridor-level volumes, not whether total driving in the system declined. The distinction matters. Removing a lane may reroute traffic or shift trip timing rather than eliminate trips, and the system-level question (does removing capacity actually reduce total VMT?) has not been directly studied.
Parking may work the same way. Building abundant free parking over decades enabled a car-dependent built environment. Removing parking requirements in new construction changes the margin of new development, but it does not restructure the existing built environment or the behavioral patterns embedded in it.
If asymmetric reversibility is a general pattern, it has profound implications for how practitioners should read the evidence. It means you cannot simply invert the sign on an observed relationship and use it as a policy forecast. “Adding X caused Y” does not license the inference “removing X will un-cause Y.” We built a highly auto-dependent, polluting, nonhuman-centered built environment over the course of a century. We now need to correct for that. But the evidence does not give us a clear path forward, and pretending otherwise does not serve the communities that need better answers.
The table
| Outcome | Parametric confidence | Range of effects | Structural uncertainty |
|---|---|---|---|
| Parking supply reduction | High | 40-53% less parking built (Seattle, Buffalo) | Low |
| Housing cost/supply | Moderate | 12-25% construction cost reduction; some rent reduction evidence | Low-moderate |
| Car ownership reduction | Low-moderate | Strong association; no causal evidence from removal | Moderate-high |
| VMT reduction | Low | 10-23% modeled decrease (cross-sectional); no end-to-end evaluation | High |
| Mode shift | Low | Indirect evidence through ownership channel | Moderate-high |
| Equity/distribution | Very low | Theoretical benefit to non-drivers through unbundling; not directly measured | Moderate |
| Emissions/environment | Very low | Not directly studied from parking reform | High |
What this means
Parking minimum elimination is primarily a housing policy with strong evidence, and a transportation policy with plausible but undemonstrated evidence. The policy clearly reduces parking oversupply and construction costs. It probably contributes to lower housing costs, though the direct evidence is thinner than the theoretical case. It may contribute to reduced car ownership and driving, but the causal chain from policy to VMT has never been measured end-to-end, and the asymmetric reversibility problem means that the evidence from decades of parking expansion cannot be straightforwardly inverted to predict the effects of parking removal.
This is a more complicated finding than either advocates or opponents typically offer. Advocates oversell the VMT reduction case by citing evidence from parking expansion as though it applies symmetrically to parking removal. Opponents dismiss the policy by pointing at neighborhood parking spillover concerns while ignoring the strong housing cost evidence. The honest assessment is that the policy is well-supported for its housing effects and speculative for its transportation effects, and that the field needs end-to-end evaluations of parking reforms in cities like Buffalo and Seattle to resolve the uncertainty.
A note on method
As with the Safe Routes to School pilot, this is a framework test, not a systematic review. I have not conducted a comprehensive literature search or applied formal inclusion/exclusion criteria. The purpose is to test whether the two-axis framework, and specifically the structural uncertainty dimension, surfaces findings that standard evidence reviews miss. The asymmetric reversibility pattern that emerged from this review is exactly the kind of structural insight the framework is designed to produce.
The evidence assessment framework is described in a companion document: Drawing Lines of Evidence: From What We Can See to What We Can Prove.
Suzanne Childress | childressssuzanne@gmail.com | unicairn.com
This post is part of an ongoing project to build systematic evidence infrastructure for transportation planning. If you are interested in collaborating, I would like to hear from you.