Who Owns the Sidewalk? Pennsylvania Moved the Delivery-Robot Market to Harrisburg
A document-first investigation of delivery-robot preemption through Baron's Four I's, comparing Pennsylvania's statewide framework with San Francisco's local permit model.
By MyAudioBooks.ai ·
Listen free: Who Owns the Sidewalk? Pennsylvania Moved the Delivery-Robot Market to Harrisburg
On July 17, 2020, the Pennsylvania Municipal League sent the House Transportation Committee a blunt warning: local governments were being denied the final decision over whether delivery robots could use their streets and sidewalks. The letter asked what would happen when a device met a wheelchair user on a narrow path, how a machine weighing as much as five hundred fifty pounds without cargo would behave in winter, and why a state application could become effective after thirty days.
Our verdict is that the market was won in the institution. The decisive business advantage was not a better wheel, sensor, or routing model. It was a statewide legal framework that converted a potentially polycentric market—many local rulemakers—from thousands of municipal negotiations into one contest in Harrisburg. Pennsylvania retained safety rules and narrow hazard authority, but it sharply limited what municipalities could regulate.
At My Audio Books dot A I, you can create your own audiobooks from prompts, turn your documents into audio, and store everything in your personal library.
Section One: Two Cities, Two Markets
San Francisco chose a local permit model. Its Board of Supervisors' 2017 record describes an ordinance requiring permits to test autonomous delivery devices on sidewalks and authorizing operational rules and penalties. The program was tightly bounded: committee amendments addressed permit duration, incident reporting, docking, site visits, and limits on the number of devices.
The following year, San Francisco set application fees at eight hundred sixty dollars for one device, one thousand five hundred forty dollars for two, and one thousand nine hundred ninety-five dollars for three. A permit decision could be appealed to the Board of Supervisors, but the device could not operate during the appeal.
That design made each deployment a local administrative event. The city controlled the testing gate, fleet size, operational conditions, appeals, and price of entry.
Pennsylvania chose a different institution. Senate Bill 1199 created a statewide authorization system under the Department of Transportation, defined operational phases, required a general plan, and established a state-controlled path from closely accompanied testing to remote monitoring. The statute did not eliminate municipalities, but it placed them inside a state framework rather than allowing each to design a separate market.
Section Two: The Sentence That Changed the Bargaining Power
Section 8517 is the commercial hinge. A municipality may permit devices on certain roads posted above twenty-five but no more than thirty-five miles per hour. It may prohibit operation on a roadway or pedestrian area when, after consulting the authorized entity, it determines that operation would constitute a hazard.
Then comes the preemption: except for those powers, a municipality may not regulate operation of a personal delivery device in pedestrian areas, roadways, shoulders, or berms under its jurisdiction. Section 8522 says regulation is governed exclusively by the state subchapter or by a municipal ordinance adopted under the narrow authority in Section 8517.
The law also requires an operator to notify a municipality at least thirty days before starting operations. Notice is not consent. Consultation is not a veto. That distinction moved bargaining power away from city halls.
In nonmarket strategy, the institution is often dispositive—decisive of the outcome. A company that loses an argument before one city council can face a ban or a restrictive pilot. A statewide rule changes the default: local officials must fit their response inside the statutory exceptions.
Section Three: Preemption Was Not Deregulation
The statute did not create a robot free-for-all. It treated personal delivery devices as pedestrians for much of the vehicle code while creating a separate authorization regime. Devices could weigh up to five hundred fifty pounds without cargo. They were limited to twelve miles per hour in pedestrian areas and twenty-five miles per hour on roadways, shoulders, or berms.
The law required dimensions no greater than thirty-two inches wide, forty-two inches long, and seventy-two inches high. Devices needed brakes, forward and rear lighting, identifying marks, a unique number, and a telephone number for reporting violations, accidents, or claims. Operators needed at least one hundred thousand dollars in liability coverage per incident.
Initial Phase One operation generally required a remote operator within thirty feet and line of sight for between ninety and one hundred eighty days, unless Penn D O T approved an exception. Phase Two allowed remote monitoring after the state determined the entity had demonstrated safe operations.
This is a regulatory floor. The business value came from making that floor statewide and limiting additional local layers.
The application itself shows how much authority moved upward. Applicants had to identify intended municipalities and routes, describe operator training, list device models, explain how emergency responders could stop the machine, identify cargo, provide insurance, outline maintenance, and submit accident procedures. They also had to attest that operations would stop or narrow during weather emergencies or other hazardous events identified by Penn D O T or a municipality.
But the state clock favoured deployment. If Penn D O T did not act on a completed authorization application within thirty days, the application was deemed approved. Much of the submitted application material was shielded from Pennsylvania's Right-to-Know Law, although route, model, insurance, identification, and accident information had specified exceptions.
That architecture produced a single compliance file and a single principal regulator. It also meant that a municipality could know a device was coming without possessing the full discretion or full record it would have held under a local permit system.
Section Four: The Opposition Understood the Strategy
The municipal coalition did not oppose technology as such. Its testimony said it could see the benefit of delivery devices, but opposed local preemption and argued that only municipalities knew whether their sidewalks and streets were prepared.
Penn D O T's own representative made a similar institutional objection. He said the proposed legislation raised safety concerns, allowed unintended operations, and overly restricted local authorities' ability to control operations. Penn D O T supported treating the devices like pedestrians, but wanted flexibility for temporary restrictions, emergencies, winter conditions, and local infrastructure.
Those objections were about subsidiarity—the principle that decisions should sit with the lowest level of government competent to make them. Snow, sidewalk width, sight distance, construction, disability access, and road geometry are local facts. A statewide rule can provide consistency while discarding local information.
The enacted framework compromised, but asymmetrically. Municipalities could identify hazards and enforce general law, yet broad hours-of-operation or zone restrictions outside the statutory exceptions were precluded. The state set the market's default; local government carried much of the street-level accountability.
At My Audio Books dot A I, you can listen to this investigation and thousands of others that explain the hidden systems behind the headlines.
Section Five: Baron's Four I's on a Sidewalk
David Baron's nonmarket framework separates Issues, Institutions, Interests, and Information. Pennsylvania's record shows all four working at once.
The issue was framed as contactless delivery, innovation, jobs, accessibility, congestion, and safety during the pandemic. At the hearing, Representative Meghan Schroeder described robots as tools for food, medicine, groceries, and other essential goods. Penn D O T answered with braking, incident reporting, weather, and operational-design concerns.
The institution was the critical choice. San Francisco's institution was a city permit and appeal process. Pennsylvania's was a state authorization with constrained municipal authority. Identical hardware entered very different markets because the legal venue assigned different vetoes.
The interests were visible in the hearing record: technology and logistics companies, state transportation officials, insurers, municipalities, townships, labour representatives, disability advocates, and residents. We should not invent a hidden coalition. The official transcript already shows an open contest among groups with different economic and safety exposures.
The information problem was the hardest. Firms knew their software, remote monitoring, fleet design, and incident data. Municipalities knew their snow routes, curb cuts, sidewalk bottlenecks, and construction cycles. The statute centralized applications and safety oversight at Penn D O T while requiring notice and consultation locally. It did not make those two information sets equal.
Section Six: The Best Case for One State Rule
The strongest case against our reading is that a city-by-city patchwork can kill a network before it starts. A delivery operator cannot efficiently design one machine for different weight limits, lighting standards, insurance rules, operating hours, licenses, and cargo restrictions in every municipality. Uniform rules reduce compliance cost, make insurance easier, and let safety learning travel across jurisdictions.
That argument wins substantial ground. Technical design, identification, braking, insurance, and data reporting benefit from a common floor. A municipality should not require a different lamp colour or mechanical architecture simply because a device crosses a boundary.
The case weakens when uniformity reaches operational context. A rule suitable for a wide suburban path may be unsuitable for a crowded historic sidewalk. Temporary snow restrictions are different from device-manufacturing standards. The law's hazard exception acknowledges that difference, but places consultation and formal determination ahead of local prohibition.
A better division would preserve statewide product and insurance standards while giving municipalities clearer authority over time, place, density, and emergency operations. That is not a rejection of scale. It is a recognition that networks operate on physical streets rather than abstract maps.
The comparison with San Francisco clarifies the trade. San Francisco's process imposed transaction costs before deployment: fees, hearings, operational restrictions, reporting, and appeal risk. Pennsylvania reduced those repeated entry costs by standardizing the market, but transferred more of the experimentation risk into statewide rules and after-the-fact hazard decisions.
Neither model is costless. Local permitting can become a veto maze in which identical safety questions are relitigated city by city. State preemption can freeze an immature standard across communities whose sidewalks, weather, disability access, and enforcement capacity differ sharply. The strategic question is not whether regulation costs money. It is who pays for learning and when.
A firm naturally prefers to pay the fixed cost of one statewide proceeding rather than the variable cost of repeated local proceedings. Residents may prefer the opposite when the physical externality sits outside their front door. Good policy has to price both forms of friction rather than treating one as innovation and the other as obstruction.
Section Seven: The Nonmarket Shortcut
For executives, the lesson is not "lobby the state." It is more exact: map which institution can grant a scalable default, which can block deployment, and which possesses the operational information you will need after the law passes.
First, distinguish product rules from use rules. Statewide design standards can reduce fragmentation. Local time-and-place rules can absorb physical differences. Conflating them invites opposition.
Second, calculate the veto map before committing capital. In San Francisco, the permit authority, Public Works Director, Board of Supervisors, and appeal process all mattered. In Pennsylvania, Penn D O T authorization and the General Assembly's statutory limits changed the sequence.
Third, bring evidence before asking for preemption. Weight, stopping distance, winter performance, wheelchair clearance, remote-operator capacity, incident rates, and insurance loss data should be public enough to let communities test safety claims. Information asymmetry is not a durable political advantage.
Fourth, build a coalition that includes the people bearing the street-level risk. Disability organizations, emergency responders, municipal engineers, couriers, small retailers, and residents are not decorative stakeholders. They define legitimacy after the launch announcement has passed.
Fifth, preserve an interstitial layer—a space between total local veto and total state control—where pilots can generate evidence. San Francisco's regime was restrictive, but it treated deployment as a test whose incidents and complaints could alter the permit. Pennsylvania's phased authorization performs part of that function at state level.
Section Eight: The Price of Winning the Venue
Preemption can create a market faster. It can also relocate accountability. When a robot blocks a curb cut, meets a child on a narrow sidewalk, or cannot brake on snow, the resident calls the municipality even if the municipality lacked broad power to design the rules.
Pennsylvania partly addressed that tension with notice, consultation, identification, insurance, incident reporting, and hazard authority. Those safeguards matter. They do not erase the institutional choice embedded in the law.
Our read is that the market was won in the institution because the statute changed who had to ask whom. Firms sought one state authorization. Municipalities received notice and bounded powers. The commercial gain was not deregulation; it was a scalable default.
At My Audio Books dot A I, you can create fiction, non-fiction, and audio from your own documents, all in one library. Learn more today at My Audio Books dot A I.
A delivery robot is visible. The nonmarket strategy underneath it is not. By the time the machine reaches the sidewalk, the decisive contest may already have occurred in a committee room far away. The price of a statewide market is that local knowledge becomes an exception rather than the rule. The market was won in the institution—and the sidewalk inherited the consequences.