Do Bike Lanes Hurt Local Businesses? What Controlled Studies Actually Find

TL;DR;

• The best North American evidence finds bike lanes usually have positive or statistically insignificant effects on nearby retail and food-service businesses.¹²
• Sales, jobs, vacancies, visits, and opinions measure different things; no single metric settles the question.
• On Toronto’s Bloor Street, customer spending rose about as much as on a control street, while cycling arrivals nearly tripled.³
• Merchants often overestimate how many customers drive, but surveys can still reveal real loading and access problems.³⁴
• Construction disruption and the completed street are different treatments and should be measured separately.⁵

The argument is usually about one parking space

When a city proposes a bike lane on a shopping street, the economic argument often begins with a vivid subtraction: this design removes 20 parking spaces. The possible additions—more people passing storefronts, safer crossings, easier short trips, more frequent visits—are harder to see before the lane exists. A curb space is concrete; a future customer is statistical.

That asymmetry helps explain why the debate can become certain long before the evidence is collected. A struggling merchant can honestly report that sales fell after a redesign. A city can honestly report that sales rose across the corridor. Both claims can be true if one business lost customers while others gained them, or if an unrelated boom lifted the street while the project slightly held it back.

The useful question is therefore not simply, “Did sales go up?” It is: what would have happened to these businesses at the same time without the bike lane? Researchers cannot rerun history, but comparison corridors, before-and-after data, difference-in-differences models, and interrupted time series can build a credible counterfactual.¹²

That is the focus here. Our broader article on cycling’s economic benefits adds household costs, health, infrastructure, and regional effects. This article stays at storefront scale: sales, employment, vacancies, customers, and the limits of claiming that a stripe of street caused any of them.

Six ledgers, six different questions

“Business impact” is not one outcome. A restaurant can serve more customers but earn less if average checks fall. Sales can rise while employment falls because the owner automates or works longer hours. A corridor can have low vacancy because landlords cut rents, or high rents because the street is thriving—but rising rents can also displace the very businesses a project was supposed to help.

The 2021 evidence review by Jamey Volker and Susan Handy found studies using sales, customer counts, reported spending, employment, vacancy, property values, openings, and owner perceptions.¹ Those measures are complementary, not interchangeable.

A convincing evaluation uses several ledgers. It should define a treatment corridor and a genuinely similar comparison before construction; collect at least a year of baseline data and preferably two or more years afterward; distinguish retail, food service, and auto-oriented businesses; and report uncertainty rather than treating every rise or fall as an effect.¹ Researchers also need to test whether the two corridors were moving in parallel before the project—the key assumption behind a difference-in-differences design.

Interrupted time-series analysis takes another route: it asks whether a corridor’s level or trend changes at the intervention date relative to its own long history. That avoids choosing a bad control street, but it needs many observations and can still mistake a simultaneous shock—a pandemic, major development, transit closure, or utility project—for a bike-lane effect.²

What the broad evidence says

Volker and Handy identified 23 North American studies that either compared spending by travel mode or measured local economic change after bicycle or pedestrian infrastructure. Fifteen impact studies covered 35 bicycle facilities, along with pedestrian and mixed projects. Ten used both before-and-after data and controls; six added statistical tests.¹

Their synthesis did not find evidence of a general retail penalty. The overall pattern was positive or statistically insignificant for retail and food-service businesses, including projects that removed parking or vehicle lanes. Among the 35 bicycle facilities, 20 had positive reported effects, 10 had unclear or insignificant effects, and five had negative reported effects. Crucially, the three studies responsible for those five negative cases had methodological limitations that prevented statistically supported conclusions about causation.¹

That is stronger than “some shops did well after a bike lane,” but weaker than “bike lanes always boost business.” The literature is still a modest collection of heterogeneous corridor studies, not a randomized experiment covering every kind of street.

Fourteen corridors, six cities, several answers

The largest coordinated U.S. effort studied 14 corridors in Portland, Seattle, San Francisco, Memphis, Minneapolis, and Indianapolis. The Portland State University team combined four possible data sources—federal employment records, state employment and wage records, retail sales tax, and proprietary establishment estimates—with trend comparisons, difference-in-differences, and interrupted time-series models.²⁶

Across corridors and methods, the researchers generally found positive or nonsignificant effects on sales and employment. Food-service sales and jobs often improved even where a driving or parking lane was reassigned.⁶ Yet the report is most valuable for showing how results can change with the dataset. Public federal employment data reach tiny geographies but include privacy-related noise; proprietary sales figures are estimates; detailed tax and wage records are accurate but often confidential. A result that appears in only one source deserves less confidence than a pattern repeated across several.²

This is also why the celebrated New York headline—retail sales rose 49% on a redesigned section of Ninth Avenue while Manhattan-wide sales rose 3%—should be treated as suggestive, not as the universal return on a protected bike lane.⁷ The city’s later economic study improved the design by comparing seven project areas with matched control corridors and borough-wide trends over multiple years.⁸ Big before-and-after numbers make good outreach; matched trends make better inference.

Bloor Street: a lane, a control, and a lesson in humility

Toronto’s Bloor Street pilot is unusually useful because researchers began planning the study before installation and used Danforth Avenue as a comparison. In 2016, the city installed 2.4 kilometers of bike lanes, removed one traffic lane and 136 on-street parking spaces, and gave researchers a politically charged test of the parking hypothesis.³⁹

The team conducted 3,005 visitor surveys and 525 merchant surveys over three collection periods, counted bicycles, and scanned ground-floor vacancies before and after installation. Its indicators pointed to a stable or improving Bloor economy:

• Merchants reporting at least 100 Saturday customers increased from 46% before installation to 62% in 2017.
• Visitors reported coming to Bloor about three more days per month after researchers adjusted for age, gender, and proximity; visit frequency did not change on Danforth.
• Bloor’s ground-floor vacancy rate stayed at 6%; Danforth’s fell from 5.2% to 3.4%.
• The share of visitors arriving by bicycle rose from 7% to 20%, while walking remained the largest mode at 48% and driving was 10%.⁹

But the honest conclusion is not that the lane caused all the gains. Spending rose at a similar rate on Bloor and Danforth, and the comparison street’s merchant-reported customer growth was stronger. The published academic analysis concluded there were no negative economic impacts associated with the lane, while the project report described the effect as positive or at least neutral because broader factors could explain growth on both streets.³⁹

Bloor also exposes the gap between merchant experience and customer behavior. Nearly half of surveyed merchants drove to the street, and a majority believed at least a quarter of customers arrived by car. Fewer than 10% of intercepted customers said they had driven. Merchants raised more concerns about business, traffic, and parking; customers were generally more positive.⁹

That mismatch does not make business owners irrational. They hear directly from a driver who cannot park but not from every nearby customer who quietly walks in. They also experience deliveries, employee commutes, and lost curb access in ways a shopper survey may miss. Their testimony is indispensable for finding a bad loading design; it is simply not a substitute for measured customer modes or transaction data.

Perception and performance can move in opposite directions

Cambridge, Massachusetts, offers an unusually candid recent example. Its 2024 study compared protected-bike-lane corridors with controls using Census employment, commercial rents and availability, proprietary spending estimates, business surveys, and customer intercepts.¹⁰ The objective datasets showed little to no consistent difference in retail employment, rents, or commercial availability. One treatment area’s availability rose slightly after installation while another’s fell.

At the same time, surveyed businesses in treatment areas were significantly more likely than control-area businesses to report declining revenue. The study could not validate those perceptions against usable sales data: the proprietary estimates contained missing observations and implausible swings of up to 100-fold. The survey covered COVID disruption, high inflation, and rising interest rates, and voluntary responses may not represent all businesses.¹⁰

The correct reading is neither “merchants were wrong” nor “the data proved harm.” It is that aggregate indicators found no systematic corridor penalty while a subset of owners reported pain that the available records could not causally assign. Our article on Cambridge’s Cycling Safety Ordinance covers the political and legal fight; the economic study shows why cities should collect baseline sales, parking, loading, and customer data before that fight begins.

The averages conceal who wins and who loses

The “local business” category combines a café serving nearby residents, a destination restaurant, a furniture store, a gas station, a repair garage, and a plumber whose vans load at the curb. A project can be neutral in aggregate while affecting those businesses differently.

A San Francisco regression study found that bicycle infrastructure and parking changes generally had no significant effect on sales after controlling for business, corridor, and neighborhood characteristics. But home-goods and auto-related businesses on bike-lane corridors did show significant sales declines. New restaurants and grocery stores had higher sales than businesses that predated the infrastructure, raising a separate possibility: street improvements may change the commercial mix rather than simply raising every incumbent’s revenue.¹¹

The broader review found similar warning signs for gas stations, auto repair and parts shops, and large home-goods stores, though the subgroup evidence is much thinner than the main retail finding.¹ These businesses may need vehicle access, bulky-goods loading, or rapid turnover at the curb in a way a bakery does not. Cities should respond with freight and access design—well-placed loading zones, accessible parking, side-street pickup, delivery windows, and enforcement—not assume an average corridor result protects every storefront.

Nor should “economic vitality” become a euphemism for displacement. Higher sales, lower vacancy, and higher commercial rents can attract investment while squeezing out legacy or low-margin tenants. Existing bike-lane studies seldom have enough time or business-level detail to separate amenity effects from broader neighborhood change.¹¹¹

Construction is not the finished bike lane

A long reconstruction project can block entrances, obscure storefronts, reroute buses, remove loading, create noise, and change traffic patterns before anyone receives the promised safer street. Those are real costs, especially for small businesses with little cash reserve.

Yet many bike-lane evaluations compare a baseline year with one or more completed-project years. They may omit the construction interval or fold it into an annual average. That answers whether the finished project changed a corridor’s trajectory; it does not prove construction was painless.

Evidence from roadway construction illustrates the distinction. A Tampa Bay study modeling establishment sales against traffic volumes and local economic conditions estimated that a two-year road-rehabilitation project reduced nearby sales by roughly 2% to 6% during construction, with somewhat larger effects for traffic-dependent businesses.⁵ A later Minnesota Department of Transportation review found the construction literature mixed and context-dependent, with effects varying by industry, customer base, project length, and access.¹² Neither result is a bike-lane estimate—and that is exactly the point. Construction disruption is a separate treatment that cities should measure and mitigate rather than attribute forever to the completed design.

Fast-build striping and posts may create little disruption; full-depth utility and streetscape work can create months of it. Project evaluations should publish the construction dates, analyze them separately, maintain signed pedestrian and delivery access, communicate schedule changes, and consider grants or other relief when verified losses are severe.

How to test a bike lane without cooking the books

A credible evaluation plan can be written before anyone knows the answer:

1. Choose comparisons in advance. Match business mix, density, transit access, rents, pre-project sales trends, and nearby development—not merely street width.
2. Record the curb. Count spaces, occupancy, turnover, loading zones, accessible spaces, illegal parking, and nearby garage capacity before and after.
3. Separate three periods. Baseline, construction, and completed operation are different exposures.
4. Measure several outcomes. Use inflation-adjusted sales, employment, openings and closures, vacancies and rents, visits, travel mode, and both merchant and customer surveys.
5. Preserve business types. Publish retail, food service, personal service, bulky-goods, and auto-oriented results where privacy and sample size allow.
6. Test the counterfactual. Plot pre-project trends, report confidence intervals, disclose missing data, and check whether results survive alternative controls and corridor boundaries.¹²¹⁰

The result may still be uncertain. A null finding can mean “no economically meaningful effect,” or it can mean the sample was too small and the data too noisy to detect one. Reports should give effect sizes and intervals, not use “not statistically significant” as a synonym for “exactly zero.”

The defensible verdict

The best available evidence does not support the prediction that reallocating parking or a travel lane for bicycles generally devastates nearby commerce. Across controlled and quasi-experimental studies, retail and food-service performance is usually positive or not detectably different from the counterfactual.¹²

The mechanism is plausible but not automatic. People walking and cycling may spend less on one visit, yet return more often; a Portland study of 78 restaurants, bars, and convenience stores found travel mode was not a significant predictor of per-trip spending after adjustment, while non-drivers visited more frequently on average.⁴ A safer, calmer street can enlarge a business’s nearby customer base, but the result depends on land use, network connections, storefront mix, loading, and execution.

So “bike lanes are good for business” is a fair summary of the average evidence only if it comes with a footnote. The more accurate claim is: well-designed bike lanes on retail streets rarely cause the broad economic damage opponents predict, often coincide with gains, and can still impose concentrated or temporary costs that good planning should detect and address. That sentence is less catchy. It is also much closer to what the studies actually find.

FAQ

Q. Do bike lanes reduce retail sales when they replace parking?
A. Controlled studies generally find positive or statistically insignificant retail effects even when projects remove parking, though auto-oriented businesses may face greater risk.¹

Q. Did Toronto’s Bloor Street bike lane help businesses?
A. Bloor’s customers, visits, and spending increased, but similar control-street growth means the safest conclusion is no negative effect, not a precisely measured windfall.³⁹

Q. Why do merchants and customer surveys disagree about driving?
A. Owners often drive themselves and hear parking complaints, while frequent nearby customers arriving on foot or bicycle create less visible friction.³⁹

Q. Can bike-lane construction hurt a small business?
A. Yes. Access and loading disruption can reduce sales during major street work, so cities should analyze construction separately from the completed lane.⁵¹²

References