What Does NYPD Enforcement Actually Buy New York?

A Cost-Benefit Audit of Low-Level Enforcement, Police Labor, Court Processing, Racial Disparities, and the Public-Safety Return on Enforcement

Executive Summary

New York City measures police activity extensively. It counts arrests, summonses, pedestrian stops, calls for service, seizures, and deployments. Those figures show what the New York City Police Department does. They do not, by themselves, establish whether those activities are efficient, effective, or worth their fiscal and social cost.

This policy analysis asks a narrower question: What does NYPD enforcement actually buy New York?

That question has become more important because several forms of low-level enforcement increased sharply after 2021. Pedestrian stops rose from 8,947 in 2021 to 25,386 in 2024. Criminal summonses increased from 45,717 to 94,542. Civil summonses increased from 11,310 to 90,354. The relevant policy inquiry is therefore not simply whether enforcement increased, but what measurable public-safety return accompanied that increase.

Criminal summonses provide a particularly useful case study. In 2024, only 9% of docketed criminal summonses ended in conviction. Sixty-nine percent were dismissed outright, another 22% were resolved through an adjournment in contemplation of dismissal, and 45% of docketed dispositions involved dismissal for legal insufficiency. When summonses rejected before docketing are included, researchers found that more than 65% of all NYPD criminal summonses issued in 2024 were facially insufficient through one mechanism or another.

Those figures do not prove that every dismissed summons lacked public value. Enforcement may produce deterrence, immediate compliance, or resolution of a community complaint without producing a conviction. But they do raise a fundamental management question: What measurable benefit justifies the police labor, supervision, court processing, administrative expense, and public burden associated with enforcement that so often fails to produce a legally sustainable result?

The analysis therefore distinguishes three concepts that are too often collapsed into one: productivity, efficiency, and effectiveness. Productivity measures how much enforcement NYPD produces. Efficiency measures what usable result is obtained from the resources consumed. Effectiveness asks whether the activity actually improves the public-safety condition it was intended to address.

The purpose of this paper is not to argue that enforcement has no value. It is to require evidence of that value. Some strategies may justify greater investment. Others may consume substantial police, court, and public resources while producing limited measurable return.

For an agency of NYPD’s size, activity alone is not enough. New York City should be able to identify what its enforcement costs, what it accomplishes, and whether the same resources could produce greater public safety if deployed differently.

I. The Problem: New York Measures Activity, Not Return

New York City has no shortage of police statistics. The New York City Police Department reports arrests, criminal summonses, civil summonses, pedestrian stops, vehicle stops, calls for service, response times, seizures, complaints, and an expanding assortment of other enforcement activity. Those numbers are useful. They tell the public where police resources are being deployed and how frequently officers are exercising governmental authority.

What they do not necessarily tell us is whether that activity worked.

That distinction matters in an agency with tens of thousands of employees, a multibillion-dollar operating budget, substantial overtime expenditures, and the ability to impose immediate legal consequences on the public. Police activity consumes officer time, supervisory time, administrative resources, court resources, and taxpayer dollars. It can also impose costs on the person subjected to the enforcement: lost time, missed work, fines, court appearances, warrants, legal expenses, and the consequences of an arrest or criminal proceeding.

The relevant policy question therefore cannot stop at how much enforcement occurred. It must examine what that enforcement produced.

A useful framework is to separate three concepts that are frequently treated as though they mean the same thing: productivity, efficiency, and effectiveness.

Productivity measures activity. How many summonses were written? How many arrests were made? How many stops were conducted? How many vehicles were seized? How many enforcement contacts occurred?

Efficiency asks a different question. What usable result was produced from the resources consumed? If thousands of officer-hours produce thousands of enforcement actions that are later rejected, dismissed, or found legally insufficient, the number of enforcement actions tells only part of the story.

Effectiveness goes further still. Did the intervention accomplish the public-safety objective that justified it? Did the condition stop? Did the location become safer? Was repeat offending reduced? Was a dangerous condition removed? Did serious crime decline? Did residents receive a lasting resolution to the problem that generated the police response?

This is not an unusual standard to impose on policing. It is increasingly central to the evidence-based policing movement itself.

In their 2025 article, “Ten Benchmarks for Evidence-Based Policing,” published by George Mason University’s Center for Evidence-Based Crime Policy in Translational Criminology, Cynthia Lum and Christopher S. Koper describe evidence-based policing as an operational approach in which police strategies, tactics, and decisions are shaped and evaluated through systematic empirical evidence. They emphasize the importance of tracking police activities, understanding how inputs and outputs are connected, and linking police decisions to actual outcomes rather than relying principally on tradition, intuition, or organizational custom.

Lum and Koper make an especially important distinction when discussing proactive patrol. The evidence, they explain, favors targeted, place-based, problem-solving, and community-engaged approaches over generalized patrol and a narrow proactive toolkit dominated by traffic and pedestrian stops. Their point is not that officers should do less. It is that police resources should be directed toward strategies with evidence that they accomplish identifiable objectives.

That framework exposes a basic weakness in the way police performance is commonly discussed in New York.

An additional 20,000 summonses may represent additional activity. It does not necessarily represent 20,000 additional units of public safety.

The same is true of arrests. An arrest is a governmental action, not a public-safety outcome. Its value depends on what the arrest accomplished, whether the charge was legally sustainable, whether prosecution followed, whether the intervention addressed an identifiable threat or recurring problem, and what resources were required to achieve that result.

There are obvious circumstances in which the benefit of enforcement may be substantial even if the eventual criminal disposition does not capture its full value. An officer may interrupt violence, remove an illegal firearm, stop a dangerous driver, disperse an immediate threat, locate a wanted person, or terminate conduct creating a serious public hazard. Those benefits should be counted.

The same analytical discipline, however, must apply when the government invokes less serious forms of enforcement. If a summons is issued for open alcohol consumption, disorderly conduct, a parks violation, public urination, noise, or another quality-of-life offense, the measure of success cannot automatically be the issuance of the summons itself. The underlying objective presumably is to change the condition that prompted government intervention.

That creates a very different performance question: did the enforcement solve the problem?

If police return repeatedly to the same location, confront the same condition, issue additional summonses, and count each summons as another unit of productivity, an activity-based system can record rising performance while the underlying problem remains unchanged.

There is also a resource-allocation problem. Police labor is finite. An officer cannot simultaneously spend an hour processing one enforcement matter and use that same hour responding to another call, investigating a robbery, following up with a shooting victim, conducting targeted traffic enforcement, developing an investigative lead, or addressing another location experiencing recurring violence.

That does not mean every officer-hour must be devoted to a felony investigation. Police departments serve broader public-order and public-service functions. It does mean that every deployment has an opportunity cost, whether the officer is working on regular time or overtime.

The public therefore needs a more complete measure of police performance. The relevant questions are not particularly complicated: What problem was the enforcement intended to address? How much police and institutional time did it consume? Did the action withstand legal review? What happened to the underlying condition? Was the result temporary or sustained? Could another intervention have accomplished the same result using fewer resources?

New York already collects much of the activity data needed to begin answering those questions. What has largely been missing is the insistence that the activity be connected to the result.

That omission has become more consequential because New York City has moved back toward substantially greater low-level enforcement.

II. The Return of Low-Level Enforcement

The scale of the change becomes clear in Justice at a Crossroads in New York City: Reexamining Crime, Enforcement, Incarceration, and Racial Disparities, the May 2025 report by Stephen Koppel, Michael Rempel, Min Xie, Olive Lu, Jeremy Travis, and Preeti Chauhan of the Data Collaborative for Justice and collaborating researchers. The report draws from more than twenty sources and traces New York City crime, enforcement, court, incarceration, and disparity measures over several decades, generally through 2023 or 2024.

The long-term picture is important because the recent increase did not begin from historically high enforcement levels. It followed decades in which many traditional enforcement measures had fallen substantially.

Pedestrian stops provide the most familiar example. NYPD-recorded stop-question-and-frisk encounters reached 685,724 in 2011. By 2021, the total had fallen to 8,947. From that low point, stops increased 184% to 25,386 in 2024.

The summons numbers moved even more dramatically.

According to the Crossroads report, civil summonses fell from 54,594 in 2018 to 10,314 in 2020 and remained at 11,310 in 2021. By 2024, NYPD issued 90,354 civil summonses—an increase of more than eightfold from 2021. More than half of the 2024 total, 50,506 summonses, involved unlawful consumption or possession of alcohol.

Criminal summonses followed a similar reversal. They had declined from 490,349 in 2006 to 45,717 in 2021, a 91% reduction. By 2024, the total had increased to 94,542. The increase did not return New York to the mass-summons levels of the mid-2000s, but it nevertheless represented more than a doubling in only three years.

Arrests also increased after the extraordinary pandemic-era decline. The Crossroads researchers report that overall arrests fell to 119,100 in 2020 and then increased 55% by 2023, reaching 184,219—slightly above the 2019 total. The composition matters: misdemeanor and nonviolent felony arrests largely returned to their 2019 levels, while violent felony arrests were 29% higher in 2023 than in 2019.

These categories should not be treated as interchangeable. A pedestrian stop, a civil summons, a criminal summons, a misdemeanor arrest, and a violent felony arrest involve different legal predicates, different institutional costs, different consequences, and potentially very different public-safety benefits. Combining them into one undifferentiated enforcement figure would conceal more than it explains.

But the direction of movement is unmistakable. After 2021, New York City substantially expanded several forms of police enforcement.

The question is what to make of that increase.

The Crossroads researchers appropriately resist making causal claims that the underlying data cannot support. Their report is expressly descriptive. They note that recent shifts in enforcement occurred alongside changes in crime and broader social conditions, making it difficult to determine how much of the movement resulted from policy choices as opposed to changing crime patterns and other forces.

That limitation is not a weakness in the analysis. It identifies the very problem a cost-benefit audit must confront.

If the City chooses to increase a particular form of enforcement, it should be able to identify what it expected that enforcement to accomplish and then evaluate whether the expected outcome occurred. Simply placing an enforcement curve beside a crime curve does not answer that question.

Suppose pedestrian stops triple while shootings decline. That sequence alone cannot establish that the stops caused the decline. Crime can move because of demographic changes, economic conditions, changes in offending patterns, prosecution, incarceration, community interventions, technology, private security, environmental changes, or a host of other factors.

The opposite inference is equally unreliable. A decline in enforcement occurring alongside a decline in crime does not prove that enforcement was unnecessary.

A serious policy analysis therefore has to move from correlation to intervention-specific evidence.

That means looking at individual enforcement strategies and asking what they were designed to accomplish. Open-container enforcement should be evaluated against the conditions it is supposed to address. Transit deployments should be evaluated against transit safety. Traffic enforcement should be evaluated against crashes, fatalities, dangerous driving, and compliance. Violence-reduction strategies should be evaluated against shootings, serious assaults, repeat offending, clearance rates, and other relevant outcomes.

The recent expansion also creates an unusual opportunity for analysis because 2021 provides a low-enforcement reference point. It is not a controlled experiment. Too many other things changed during and after the pandemic to treat it as one. But the size of the enforcement movement makes it possible to ask a much more precise question than whether policing in the abstract “works.”

Between 2021 and 2024, New York City purchased substantially more enforcement activity.

What did the additional activity produce?

The best place to begin is not with a theoretical debate over policing. It is with an enforcement system where the available data allows us to follow thousands of police actions from issuance into the courts.

That system is criminal summons enforcement.

III. Criminal and Civil Summonses: A Case Study in Enforcement Yield

Criminal Summonses: Following the Enforcement Through the System

A summons presents an unusually useful test of enforcement performance because it creates a record that can be followed beyond the street encounter. The police issue the instrument. The court determines whether it can be docketed. If it proceeds, the case produces a disposition. That makes it possible to distinguish the number of enforcement actions generated from the number that survive basic legal and administrative review.

The most recent detailed examination is Anna Stenkamp and Michael Rempel’s September 2025 Data Collaborative for Justice research brief, Trends in Criminal Summons Issuance, Outcomes, and Racial Disparities, 2023–2024. The study uses data supplied by the New York State Office of Court Administration and follows NYPD criminal summonses through the court system.

The authors make an important methodological distinction from the outset. Their court-based dataset is not identical to the NYPD public dataset used in the Crossroads report. Public NYPD summons data produces similar trendlines but higher overall totals. Stenkamp and Rempel attribute the difference to the court-based source and exclusions used in their analysis, including the omission of summonses issued by agencies other than NYPD and summonses issued to businesses. The court data, in turn, permits them to distinguish defective summonses and track how docketed cases were ultimately resolved.

That explains an apparent numerical difference that should not be obscured. The Crossroads report records 94,542 criminal summonses in 2024 using NYPD public data. Stenkamp and Rempel identify 78,848 summonses in their 2024 court-based series. The sources are measuring the system differently. The relevant point for the present analysis is that both show the same broad post-2021 increase, while the court dataset allows a closer examination of what happened after issuance.

The first attrition occurs before a defendant ever reaches a judge.

In 2024, courts rejected 15,908 NYPD summonses as defective before docketing. That represented 20% of the summonses in the court-based dataset, compared with only 4% in 2013. The defect rate therefore increased fivefold over the longer period, with much of the increase occurring after 2021.

Some of those defects were strikingly basic. In 2024, “original summons not being on file” accounted for 26% of defective summonses. The report explains that this occurs when the wrong copy of the summons form is delivered to the court; the court cannot scan the incorrect copy and rejects it.

The summonses that survived that initial screening did not perform much better once docketed.

In 2024, only 9% of docketed criminal summonses resulted in a conviction. Sixty-nine percent were dismissed outright. Another 22% ended in an adjournment in contemplation of dismissal, which the authors note ordinarily culminates in dismissal if no new summons is received during the applicable period. Criminal-summons conviction rates had remained low throughout the preceding years, ranging from 4% to 13% since 2019.

More revealing is the reason for many of the dismissals.

Forty-five percent of all docketed summons dispositions in 2024 involved dismissal for legal insufficiency. The report identifies examples including illegible handwriting and an incorrect penal provision. The legal-insufficiency rate had been 35% in 2019, declined to 28% in 2021, and then rose to 45% by 2024. Stenkamp and Rempel state that this increase suggests a possible emphasis on quantity over accuracy.

When the two stages are combined—the summonses rejected as defective before docketing and those dismissed as legally insufficient after reaching court—the result is extraordinary. More than 65% of all summonses issued in the 2024 court-based dataset were found facially insufficient through one of those two mechanisms. The Data Collaborative for Justice’s public summary of the study reports the same finding.

That figure should not be distorted. Facial insufficiency is not the same thing as a judicial finding that the underlying conduct did not occur. Nor does a dismissal necessarily establish that the original police intervention produced no immediate benefit. A person may have stopped the complained-of conduct after receiving the summons. The enforcement may have responded to a legitimate community concern. The encounter itself may have deterred repetition.

But those possibilities do not answer the efficiency question created by the data.

If the government chooses a criminal enforcement mechanism, prepares an accusatory instrument, requires a person to appear in criminal court, and activates the machinery necessary to process that case, legal sufficiency is not an incidental consideration. It is a minimum condition for the enforcement action to function as designed.

A system in which more than six out of every ten issued instruments fail that threshold is consuming resources before the ultimate policy benefit is even considered.

The cost does not end with the issuing officer. There is supervisory oversight, transmission and processing, court intake, clerical work, courtroom time, judicial review, and the time imposed on the recipient. When an instrument is rejected because the wrong copy was submitted, an incorrect statute was written, or the accusation was legally insufficient, those resources have already been consumed.

That is why summons volume is an incomplete performance measure.

The more useful measures would include the cost per summons issued, the cost per legally sufficient summons, the cost per sustainable disposition, and—most importantly—the cost per demonstrated improvement in the condition the enforcement was intended to address.

Those are materially different calculations.

An agency can increase the first number while performing worse on the remaining three.

The Return to Criminal Enforcement for Conduct the City Had Already Diverted

The summons data also raises a second policy question: why use the more burdensome criminal process when the law expressly created a civil alternative?

New York City’s Criminal Justice Reform Act, effective in 2017, made five categories of low-level misconduct eligible for civil rather than criminal summons treatment: public consumption of alcohol, public urination, parks offenses, noise, and littering or spitting. Civil summonses are adjudicated through the Office of Administrative Trials and Hearings (OATH) and do not carry the same warrant, incarceration, or criminal-record exposure associated with the criminal summons process. Police nevertheless retain authority under specified circumstances to issue a criminal summons instead. Stephen Koppel and Anna Stenkamp explain that structure in their April 2024 Data Collaborative for Justice report, Racial Disparities in New York City Civil Summonses, 2019–2022.

Initially, the statutory diversion worked largely as intended. Stenkamp and Rempel report that docketed criminal summonses for CJRA-eligible offenses declined 94% during the first year after implementation.

That changed after 2021.

Criminal summonses for open-container offenses increased from 2,010 in 2021 to 20,595 in 2023—an increase of 924%. CJRA-eligible offenses, which had represented only 8% to 12% of criminal summonses from 2018 through 2021, increased to 37% in 2023. That year, 27,884 criminal summonses involved conduct for which a civil enforcement mechanism was legally available.

The movement partially reversed in 2024. CJRA-eligible offenses declined to 14% of criminal summonses, while open-container cases fell from 27% of the 2023 criminal-summons total to 9% in 2024.

That rapid swing itself deserves attention. If the underlying conduct did not change comparably, the numbers suggest how heavily summons outcomes may depend on enforcement policy and officer discretion rather than simply the prevalence of particular behavior. The existing studies do not establish why those choices changed, and they should not be made to prove more than they do. They do establish that the same category of conduct can be routed very differently through the City’s enforcement systems from one period to another.

A cost-benefit analysis should therefore compare not merely enforcement against no enforcement, but one enforcement mechanism against another.

If a civil summons can address the same condition without requiring criminal-court processing, warrant exposure, or potential criminal penalties, the City should be able to explain when the additional cost of the criminal process produces an additional public-safety benefit.

Civil Summonses: Less Punitive, But Not Costless

The civil system was designed in part to reduce the collateral consequences associated with low-level criminal enforcement. That is an important benefit.

It does not make civil enforcement free.

Koppel and Stenkamp reported that NYPD issued 27,673 civil summonses in 2022, nearly four times the 7,252 issued in 2021. Eighty-three percent involved public consumption of alcohol. The City’s total fines associated with civil summonses that year were $443,879, and 84% of the fines imposed were $25 or less.

The later Crossroads report shows how much further the system expanded: 90,354 civil summonses in 2024, including 50,506 for unlawful consumption or possession of alcohol.

That growth creates a separate efficiency inquiry.

A civil summons requires an officer to observe or investigate the conduct, interact with the person, prepare the instrument, and complete the corresponding administrative work. The case then enters OATH’s administrative system. Notices must be processed. Hearings may occur. Decisions must be issued. Fines may have to be collected. Failures to respond require additional administrative handling.

Those costs may be justified. But justification requires an outcome beyond the number of summonses produced.

If the purpose of open-container enforcement is to reduce drinking-related disorder at a particular location, the appropriate measure is not simply how many summonses officers wrote there. It is whether the condition improved, how long the improvement lasted, whether complaints declined, whether police had to return repeatedly, and whether another intervention could have achieved the same result at lower cost.

That is the distinction that should govern the remainder of the analysis.

Police enforcement has value when it produces public value. The next task is to determine what New York City actually spends to produce it.

IV. What Enforcement Actually Costs

The fiscal cost of enforcement is more difficult to identify than the number of enforcement actions because New York City does not account for policing on an encounter-by-encounter basis. There is no published line item showing what it costs to issue a criminal summons, conduct a pedestrian stop, make a misdemeanor arrest, or process a quality-of-life complaint from initial response through final disposition.

That does not mean the cost cannot be measured. It means the relevant expenses are spread across several budgets and accounting systems.

Police Labor Is the Largest Cost

The first component is officer time. NYPD personnel must respond to the condition, conduct the encounter, determine what enforcement action to take, prepare the necessary paperwork, obtain supervisory review where required, and complete whatever follow-up the action generates. An arrest can add transportation, prisoner processing, property handling, complaint preparation, interaction with prosecutors, and overtime. A summons generally requires less time, but it still requires paid police labor.

NYC Open Data’s Citywide Payroll Data, maintained by the New York City Office of Payroll Administration, provides actual regular compensation, regular hours, overtime hours, overtime pay, base salary, and other compensation for City employees by fiscal year. That dataset makes it possible to move beyond a generic estimate of an “officer hourly rate” and examine what New York actually pays employees occupying different NYPD titles.

Overtime is particularly significant. In its June 2026 report, A Framework for Structural NYPD Overtime Reform, the Office of the New York City Comptroller reported that NYPD accounts for approximately 40% of the City’s annual overtime cost. Uniformed NYPD overtime reached approximately $960 million in Fiscal Year 2025, and the Comptroller projected approximately $890 million for Fiscal Year 2026—still one of the highest annual totals on record.

The composition of that spending is important. The Comptroller’s Annual State of the City’s Economy and Finances 2025 reported approximately $168 million in investigative overtime in Fiscal Year 2025, $161 million for transit safety, $82 million associated with new arrests, $69 million for crime reduction, and $259 million for events, along with operational and other categories.

Those categories do not permit a conclusion that a particular summons costs a particular amount. Nor should every dollar of “new arrest” overtime be treated as the cost of low-level policing. The figures establish something more basic: substantial portions of NYPD’s labor expenditures can be connected to identifiable operational functions, and those expenditures are large enough that the City should be asking what return they produce.

There is also a difference between payroll cost and the full cost of employing an officer.

The New York City Independent Budget Office demonstrated that point in its 2021 report, A Full Accounting: How Much Does New York City Spend on Its Criminal Justice System? IBO allocated centrally budgeted pension and fringe-benefit expenses back to the agencies that generated them. For NYPD, adding pension and fringe expenses increased the Department’s Fiscal Year 2021 cost by approximately 76% over its direct agency budget; adding allocated debt service increased the total by approximately 80%.

Those historical percentages should not simply be applied to current NYPD spending as though they were permanent multipliers. Pension contributions, fringe costs, debt obligations, staffing, and compensation change over time. The importance of the IBO analysis is methodological: the taxpayer cost of an officer-hour is materially greater than the salary appearing on the officer’s paycheck.

A defensible cost analysis should therefore distinguish between direct cash compensation and a fully loaded personnel cost that includes the City’s attributable fringe and pension obligations.

Enforcement Continues to Cost Money After the Street Encounter Ends

Police labor is only the beginning.

The Mayor’s Office of Management and Budget’s FY 2026 Executive Budget Function Analysis provides one example. It identifies an NYPD Criminal Justice Bureau whose responsibilities include borough arrest processing, pre-arraignment processing, prisoner control, transportation, and detention. For Fiscal Year 2026, OMB budgeted approximately $70.5 million for that function, including 185 uniformed and 187 civilian positions.

Not every dollar in that budget can be attributed to low-level arrests, and criminal summonses ordinarily do not generate the same prisoner-processing costs. But the budget demonstrates why counting only the time of the arresting officer understates the cost of arrest-based enforcement. Once police elect to take someone into custody, an institutional processing system activates behind the officer.

Civil summonses activate a different system.

The New York City Council Finance Division’s Fiscal Year 2026 analysis of the Office of Administrative Trials and Hearings, prepared by Senior Financial Analyst Saiyeumul Hamid, reported an OATH budget of approximately $81.6 million and 470 budgeted full-time positions. The financial plan included additional funding specifically for increased summons-related postage, remote hearing services used by the Hearings Division, and Clerk’s Office staffing.

OATH’s own description of its Hearings Division illustrates the work generated by a summons after issuance. Clerks schedule hearings, process rescheduling requests, send decisions and notices, check in respondents, process penalties and payments, issue refunds, and operate the call center.

Again, it would be incorrect to divide OATH’s entire budget by the number of NYPD civil summonses. OATH adjudicates summonses issued by numerous City agencies and performs functions unrelated to NYPD quality-of-life enforcement. But the administrative burden is real, and at least part of it is attributable to the enforcement decision that placed the matter into the system.

Criminal summonses raise the same problem on the state-court side. The New York State Unified Court System’s Fiscal Year 2026 Judiciary Budget sought approximately $361.5 million for the broader City and District Courts program, which includes but is not limited to New York City Criminal Court. The Unified Court System’s 2025 Annual Report separately recorded 336,093 filings in the Criminal Court of the City of New York, a category that includes arrests, certain tickets, and summons cases.

The published budget does not isolate the precise judicial cost attributable to NYPD criminal summonses. That is an accounting limitation, not evidence that the cost is zero.

A criminal summons rejected as defective still had to be transmitted and reviewed. A summons docketed and later dismissed for legal insufficiency still required clerical handling and judicial action. A respondent who appeared in court consumed courtroom capacity even if the case lasted only minutes. Multiply those events across tens of thousands of cases, and seemingly minor processing costs become material.

This is especially relevant to the 2024 criminal-summons data. Stenkamp and Rempel found 15,908 summonses rejected before docketing as defective. Of the cases that proceeded, 45% of dispositions involved legal-insufficiency dismissals.

Those are not merely disposition statistics. They represent consumed government capacity.

Opportunity Cost Is a Real Cost Even When No Overtime Is Paid

One of the most important costs never appears as a separate line in a City budget.

If two officers spend part of a tour on an enforcement matter, the City has already purchased their time regardless of whether the encounter generates overtime. But those officers cannot use the same time for another purpose.

That is opportunity cost.

The concept is particularly important in policing because demands on police personnel are not interchangeable. An hour spent writing and processing a summons is an hour not available to answer another call, conduct a follow-up investigation, locate a witness, perform targeted traffic enforcement, investigate a shooting, work a pattern robbery, or address another recurring condition.

The relevant comparison is therefore not always enforcement versus doing nothing.

It may be one enforcement activity versus another use of the same police resource.

Lum and Koper’s evidence-based-policing framework is particularly useful here. Their discussion of proactive policing emphasizes targeted, place-based, problem-solving strategies rather than generalized activity, and their tracking benchmark expressly calls for agencies to understand patrol inputs, outputs, and the relationship between them. In their formulation, commanders cannot strategically deploy personnel if they do not know what officers are doing with their time and how those activities relate to outcomes.

That principle has obvious fiscal implications for NYPD.

Suppose one enforcement strategy requires large numbers of police contacts, extensive paperwork, repeat deployments to the same locations, and substantial overtime but produces only temporary compliance. A second strategy might require fewer contacts but resolve the underlying condition for a longer period. Even if both strategies technically enforce the same law, their cost-effectiveness is not equivalent.

That is why the City ultimately needs something it does not currently publish in a usable form: average officer and supervisor time by enforcement activity.

Without that information, it is difficult to calculate the marginal cost of a criminal summons, civil summons, pedestrian stop, vehicle stop, misdemeanor arrest, or felony arrest. The missing data should not prevent analysis. It should become part of the policy recommendation.

A police department capable of recording enormous quantities of enforcement data should also be capable of determining approximately how many personnel-hours different enforcement strategies consume.

Error and Liability Belong on the Same Ledger

There is one more category of cost that conventional productivity statistics largely ignore: what happens when the police action itself generates a complaint, investigation, claim, or lawsuit.

The Office of the New York City Comptroller’s September 2025 report, A Blueprint for Department-Wide Restraint: An Analysis of NYPD Excessive Force Complaints, Claims, and Lawsuits, reported that 6,082 Police Action claims were filed in Fiscal Year 2025 and that the City paid more than $113 million in settlements associated with Police Action claims.

Those figures cannot fairly be assigned to low-level enforcement. Police Action claims arise from many kinds of encounters, including circumstances far removed from summons activity. Treating the entire $113 million as a cost of quality-of-life policing would be analytically indefensible.

But excluding liability altogether would be equally mistaken.

Every police contact carries some risk of escalation, complaint, administrative investigation, injury, or litigation. The probability is not uniform across all encounters. A voluntary conversation, a summons, a forcible arrest, a vehicle pursuit, and the execution of a search warrant involve very different risk profiles.

The appropriate question is therefore whether the City can identify civil-liability costs by the type of encounter that generated them.

If certain categories of enforcement consistently produce higher complaint or claims rates, that information belongs in the resource-allocation decision. If other strategies accomplish comparable public-safety objectives with lower litigation exposure, that difference also has economic value.

The same principle applies to basic administrative error. A criminal summons rejected because the wrong copy was submitted is not merely a paperwork problem. It represents police time that has already been expended and court resources consumed without producing a viable case.

Error has a price.

The challenge is to make that price visible.

V. What Does the City Receive in Return?

A cost-benefit analysis cannot stop with cost. A police intervention that is expensive may still be justified if the public-safety return is sufficiently large.

The question is how that return should be measured.

Some benefits are straightforward. Police may stop dangerous conduct that is occurring in real time. They may remove an illegal firearm, arrest a person committing violence, stop an intoxicated driver, recover stolen property, locate a wanted suspect, clear a violent crime, or eliminate a hazardous traffic condition. In those circumstances, the connection between the intervention and the immediate result may be readily observable.

Other benefits are more difficult to measure. Quality-of-life enforcement is frequently justified in terms of deterrence, order maintenance, responsiveness to community complaints, prevention of escalation, or public perceptions of safety. Those may be legitimate governmental objectives. But because they are less direct, they require greater—not less—attention to measurement.

A summons is not itself proof of deterrence.

An arrest is not itself proof that future crime was prevented.

A pedestrian stop is not itself proof that a neighborhood became safer.

Those are enforcement events. The claimed benefits occur after or because of them.

The difference matters because New York’s recent history does not support a simple assumption that the volume of police enforcement and the level of crime move together in a predictable way.

The Crossroads executive summary reports that from 1990 to 2019, reported violent index crime in New York City declined by 77%, while property index crime declined by 81%. Shootings fell from 5,269 incidents in 1993 to 777 in 2019. At the same time, the longer Crossroads analysis documents enormous reductions in pedestrian stops, criminal summonses, drug arrests, misdemeanor arrests, and other forms of enforcement across portions of that period.

That history does not establish that lower enforcement caused lower crime. It does establish that the relationship is more complicated than an activity count.

The early 2020s illustrate the same point in the opposite direction. Violent and property crime increased after the pandemic disruption, several enforcement measures subsequently increased, and some crime measures later declined. The Crossroads researchers expressly caution that their descriptive data cannot establish which changes caused which outcomes.

A credible cost-benefit analysis must preserve that limitation.

The appropriate inquiry is intervention-specific.

If NYPD substantially increases open-container enforcement, the relevant measures should include whether alcohol-related disorder declined at the places receiving the enforcement, whether complaints declined, whether repeat enforcement became less necessary, and whether any change persisted after officers left.

If the Department increases pedestrian stops in particular locations to address weapons violence, the analysis should examine weapon recoveries, shooting trends, serious violence, the hit rate of those stops, the geographic concentration of the intervention, and whether comparable locations receiving different police strategies experienced different outcomes.

If additional transit overtime is justified as a safety intervention, the City should examine crime and victimization in the transit system, serious incidents, repeat offending, response time, ridership, and the marginal effect of the added deployment—not merely the number of officers assigned underground.

If vehicle enforcement is intended to reduce dangerous driving, the appropriate outcomes are crashes, injuries, fatalities, repeat violations, and changes in the targeted behavior.

The point is not to demand an impossible level of scientific certainty from every police action. Public agencies routinely make decisions under uncertainty. The point is to distinguish between a plausible theory of benefit and evidence that the benefit actually occurred.

Recorded Crime Is Only Part of the Benefit Calculation

Police-recorded crime is also an incomplete measure of public safety because many crimes never enter police records.

That limitation is examined directly in Justice at a Crossroads in New York City: Studying Crimes in New York City Using the National Crime Victimization Survey, a May 2025 report by Min Xie, Preeti Chauhan, Michael Rempel, and Jeremy Travis. Rather than relying exclusively on offenses reported to NYPD, the authors use the federal National Crime Victimization Survey to examine both reported and unreported victimization in New York City from 1996 through 2022.

The distinction is substantial.

The researchers found that more than half of crimes in New York City were not reported to police. Violent-crime reporting declined from approximately 49% during the late 1990s to approximately 35% near the end of the period they examined. Property-crime reporting declined from approximately 31% to 21%.

That means an enforcement policy cannot be evaluated solely through NYPD complaint statistics without recognizing that those statistics measure reported crime, not all victimization.

The NCVS findings also complicate the historical picture in a useful way. The researchers found that nonfatal violent victimization in New York City was 61% lower in 2021 than in 1997, while property victimization was 38% lower. Those findings broadly confirm the City’s long-term improvement in safety using a source independent of NYPD’s own complaint records.

At the same time, the study found that New York City’s relative advantage over other large cities had narrowed and that victim reporting had declined.

That is why the benefit side of the ledger should include more than one outcome.

A serious assessment of policing should consider reported crime, victimization, serious violence, shootings, traffic injuries, case clearances, repeat complaints, victim cooperation, response times, and the specific condition an enforcement initiative was designed to change.

No single statistic can carry the analysis.

Enforcement Should Be Compared Against the Alternative That Could Have Been Used

Cost-benefit analysis also requires a counterfactual.

The relevant question is rarely whether enforcement produces any benefit at all. Almost any intervention can be associated with some benefit. The more useful question is whether the chosen enforcement strategy produces greater net benefit than the available alternatives.

The Criminal Justice Reform Act provides a concrete example.

For certain low-level conduct, the City has already decided that either criminal or civil enforcement may be available under specified circumstances. If the immediate condition can be addressed through a civil summons, using the criminal process should require some additional justification because criminal enforcement carries additional institutional and personal costs.

The same reasoning applies more broadly.

A warning may be sufficient in some circumstances.

A civil summons may accomplish what a criminal summons would otherwise accomplish.

Problem-solving at a recurring location may outperform repeated individual enforcement.

Targeted enforcement directed at a documented high-risk place may generate more safety than diffuse low-yield activity.

Better investigations may produce a greater return than increasing street contacts.

The policy objective is not to choose the least intrusive intervention regardless of effectiveness. It is to identify the intervention that produces the greatest public-safety benefit relative to its cost and consequences.

That is the essence of evidence-based resource allocation.

The Department should therefore be able to answer a question more specific than whether an enforcement initiative produced “results.”

What result did it produce that would not likely have occurred, or would not have occurred as effectively, through a less costly alternative?

Without that comparison, enforcement volume can become self-validating: officers conduct more enforcement, the Department reports more enforcement, and the increased number is then presented as evidence that the initiative succeeded.

That is an accounting of activity, not an evaluation of policy.

VI. Who Bears the Cost of Enforcement?

Even an enforcement strategy that produces an aggregate public benefit raises a separate question: how are its burdens distributed?

Cost-benefit analysis at the citywide level can conceal that problem. A policy may produce benefits for one population while concentrating police contacts, court appearances, fines, lost time, and other costs in another. That does not, standing alone, determine whether the policy is justified. It does mean that distribution is part of the analysis rather than an afterthought.

The summons data shows why.

Stenkamp and Rempel found that Black and Hispanic New Yorkers received 78% of docketed criminal summonses in 2024 while constituting approximately 51% of the City’s general population. Black residents received criminal summonses at 6.2 times the rate of white residents, while Hispanic residents received them at 4.0 times the white rate.

Those disparities were lower than their 2021 levels, an important part of the trend that should not be omitted. The Black-white rate ratio declined from 11.7 to 6.2, and the Hispanic-white ratio declined from 6.1 to 4.0 between 2021 and 2024.

The citywide decline, however, obscures substantial differences by borough. In Queens, the Hispanic-white criminal-summons rate ratio increased from 5.5 in 2021 to 11.4 in 2024. Manhattan, by contrast, was the only borough in which both the Black-white and Hispanic-white ratios declined over that period.

That variation matters because it suggests that citywide enforcement policy does not operate uniformly across New York.

Income Does Not Eliminate the Disparity

One possible explanation for racial disparity in police enforcement is geography. Police resources and enforcement may be more concentrated in neighborhoods experiencing higher levels of disorder, poverty, calls for service, or reported crime. If those neighborhoods are also disproportionately Black or Hispanic, population-based racial differences can emerge without establishing that race caused an individual enforcement decision.

The available summons research does not permit a causal conclusion about why the disparities exist.

But the neighborhood-income analysis makes a simple poverty explanation difficult to sustain by itself.

In the 2024 criminal-summons data, Black and Hispanic residents accounted for at least 63% of docketed summonses across low-, middle-, and high-income neighborhood groupings. The disparity was particularly pronounced in neighborhoods with median household income above $125,000. There, Black residents constituted approximately 6% of the population but received 40% of docketed criminal summonses in 2024. Hispanic residents constituted approximately 10% of the population but received 24%.

The earlier civil-summons research identified a similar pattern.

In Racial Disparities in New York City Civil Summonses, 2019–2022, Koppel and Stenkamp found that the largest population-adjusted racial disparities in 2022 occurred in the most affluent neighborhood quintile. Hispanic residents were 7.5 times as likely as white residents to receive a civil summons there; Black residents were 5.5 times as likely.

The two studies examine different enforcement systems and different time periods, so their ratios should not be merged. What can fairly be said is that both independently found pronounced racial disparities in affluent areas.

That finding changes the policy inquiry.

If concentrated poverty and enforcement intensity were sufficient explanations, one might expect disparities to narrow substantially in the City’s wealthiest neighborhoods. Instead, the research found some of the widest relative differences there.

That does not prove discriminatory intent, officer bias, or unlawful enforcement. The civil-summons researchers expressly caution that population benchmarking captures disparities arising from many possible factors and cannot identify a particular cause without a more complex analysis.

The criminal-summons study contains additional limitations. Race or ethnicity information was missing in 8% of summonses in 2024, and the authors identify Census classification issues affecting population-rate calculations, including the treatment of people identifying as both Black and Hispanic.

Those qualifications matter. They keep descriptive disparity from being converted into an unsupported causal conclusion.

They do not make the disparity irrelevant.

From a public-policy perspective, the appropriate question is whether the City understands why the burdens of a particular enforcement strategy are distributed as they are and whether the distribution is connected to the public-safety objective the strategy is supposed to serve.

The Burden Is More Than the Fine

The cost imposed on the recipient of enforcement also cannot be measured solely by the amount of a ticket.

A criminal summons generally requires a court appearance and carries the possibility of a warrant if the recipient fails to appear. A civil summons is less punitive and can generally be resolved without entering the criminal system, but it still requires payment, participation in an administrative process, or attendance at a hearing if contested.

Those obligations consume time.

For an hourly worker, a court appearance can mean lost wages. For a parent or caregiver, it can require arranging childcare. Transportation has a cost. Legal advice has a cost. A missed appearance or unresolved matter can create additional administrative consequences.

The civil-summons report provides one indication that those burdens do not fall evenly. In 2022, the failure-to-appear rate for civil summonses was 64% among Black recipients, 57% among Hispanic recipients, and 31% among white recipients. The overall rate had increased from 48% in 2019 to 54% in 2022.

Failure to appear in the OATH civil system does not carry the arrest-warrant consequences associated with a criminal summons. But the size of the difference raises questions about whether the administrative process itself interacts differently with different populations—through ability to pay, access to information, work schedules, mobility, trust, or other factors the study did not determine.

That is another reason why the cost-benefit inquiry should extend beyond government expenditure.

The City pays to enforce the law. The person subjected to enforcement also pays, sometimes in money and sometimes in time.

Public Trust Has Operational Value

There is a final distributional cost that is more difficult to price but directly connected to police effectiveness.

The NCVS study found that victims increasingly identified a lack of confidence in police assistance as a reason for not reporting crime. Among victims of violent crime who did not contact police, the share saying that “Police wouldn’t help” increased from 7% in 1997 to 16% in 2021. Among property-crime victims, the corresponding figure increased from 12% to 18%.

Those figures do not establish that any particular NYPD enforcement practice caused the decline in confidence. The data should not be stretched that far.

They establish something else that is important to the cost-benefit analysis: public confidence has operational consequences.

Police agencies depend on civilians to report crimes, identify suspects, provide video, appear as witnesses, supply intelligence, cooperate with investigators, and call when dangerous conditions develop. A police department can possess enormous formal authority and still lose investigative capacity if the public becomes less willing to use it.

That makes legitimacy more than a public-relations concern.

If an enforcement practice produces modest public-order benefits but substantially damages cooperation in the same community, the loss of information and willingness to report crime is part of the policy cost. Conversely, an enforcement strategy that improves both safety and public confidence generates a benefit beyond whatever arrest or summons statistics it produces.

The City should therefore assess enforcement distribution not merely as a demographic reporting requirement, but as part of operational performance.

The question is not simply who received the summons or who was stopped.

It is whether the burdens imposed by the policy are connected to its demonstrated public-safety benefits—and whether the enforcement strategy strengthens or weakens the public cooperation upon which effective policing ultimately depends.

VII. A Better Accountability Model for NYPD

The data examined throughout this analysis points to a basic measurement problem. New York City can describe how much enforcement NYPD produces with considerable precision. It can report how many summonses were issued, how many arrests were made, how many pedestrian stops occurred, how much overtime was spent, and how many cases entered the courts. What remains much less developed is a system for connecting those activities to cost, legal sustainability, and measurable public-safety outcomes.

A more complete accountability model would begin by changing the unit of measurement.

The relevant question is not simply how many enforcement actions occurred. It is what happened because they occurred.

From Enforcement Volume to Enforcement Return

Traditional activity measures have an obvious administrative value. A police department needs to know what its officers are doing. Supervisors need to know whether commands are active, whether enforcement is occurring in designated areas, whether officers are responding to identified conditions, and whether resources are being deployed as directed.

The problem arises when the activity measure becomes the performance measure.

A command that issues 5,000 summonses has unquestionably produced more summonses than a command that issues 2,500. That comparison says nothing about whether the first command reduced the underlying problem more effectively, whether its summonses were legally sufficient, whether the same locations repeatedly generated enforcement, whether complaints declined, or whether twice the enforcement produced twice the public benefit.

The distinction is particularly important in light of the criminal-summons data. An enforcement system can generate substantial activity while producing a much smaller number of legally sustainable cases. The 2024 findings discussed earlier—high pre-docket defect rates, substantial legal-insufficiency dismissals, and a conviction rate below 10%—demonstrate why the number of instruments issued cannot function as a sufficient measure of success.

The better measure is enforcement return.

That concept does not require every police action to result in a conviction. A successful intervention may stop dangerous conduct, resolve a recurring condition, deter repetition, recover contraband, protect a victim, or eliminate a hazardous situation without producing a criminal judgment. But the Department would have to identify which outcome the intervention was intended to achieve and whether that outcome occurred.

That change sounds modest. Institutionally, it is significant.

It moves the performance question from “How much did officers do?” to “What did the activity accomplish?”

The City Already Collects Much of the Necessary Information

A return-based model would not require New York to create an entirely new statistical infrastructure. Much of the information already exists in separate systems.

NYPD records enforcement activity. Payroll records capture regular and overtime compensation. Budget documents identify personnel and operational expenditures. OATH processes civil summonses. The court system records criminal summons filings and dispositions. Claims data captures at least some of the financial consequences of police encounters. Crime statistics, calls for service, 311 complaints, traffic data, and victimization research provide different measures of the conditions enforcement is intended to address.

The present problem is fragmentation.

Those systems answer different institutional questions and generally do not produce a single measure showing the relationship between police activity, resources consumed, adjudicative outcome, and public-safety effect.

An enforcement-return framework would connect them.

For each significant enforcement category, the basic record would include the number of actions taken, the police labor associated with those actions, the direct and fully loaded personnel cost, overtime attributable to the activity where identifiable, the downstream processing cost, the legal disposition, and the observable change in the condition the enforcement was intended to address.

That framework could be applied separately to criminal summonses, civil summonses, pedestrian stops, vehicle stops, misdemeanor arrests, traffic initiatives, transit deployments, and other large enforcement programs.

The categories should remain separate because their purposes and costs differ. A pedestrian stop directed at weapons enforcement should not be evaluated under the same outcome measure as a civil summons for public alcohol consumption. A traffic initiative should be tested against traffic harm. A transit deployment should be tested against conditions in the transit system. A quality-of-life initiative should be evaluated against the quality-of-life problem that triggered it.

The common principle is that the outcome measure must correspond to the stated purpose of the enforcement.

Measure Whether the Problem Was Solved, Not Merely Whether Police Responded

One of the most useful changes would be to track recurrence.

Consider a location generating repeated complaints about the same condition. NYPD responds on Monday and issues several summonses. Officers return on Wednesday and issue more. The following weekend produces another deployment and another group of summonses.

Under an activity-based system, each additional enforcement action increases the recorded output.

Under an outcome-based system, repeated enforcement may signal that the original intervention did not resolve the problem.

That distinction is particularly important for quality-of-life policing because many of the underlying conditions are place-based and recurring. Noise, disorder, illegal vehicle activity, public drinking, unsafe traffic behavior, and other complaints frequently arise from identifiable locations or patterns rather than isolated individuals.

The useful performance measure is therefore not simply how many times police acted.

It is whether police intervention reduced the need to act again.

A return-based system could compare enforcement activity with repeat 311 complaints, repeat calls for service, repeat summons activity at the same location, recurring arrests, or other appropriate indicators. A location requiring fifty summonses over several months might represent aggressive enforcement. It might also represent an unresolved problem.

The distinction cannot be determined from the summons count alone.

This is where evidence-based policing becomes particularly relevant. The framework described earlier by Lum and Koper emphasizes tracking police activity precisely so that agencies can connect deployment to outcomes and adjust strategies when interventions do not perform as intended. The operational value of tracking is not the production of more statistics. It is the ability to change tactics when the evidence shows that the existing approach is not solving the problem.

Cost Has to Become Part of Performance Measurement

A second change would be to incorporate resource consumption into routine evaluation.

Two strategies can produce the same apparent result while requiring very different amounts of police labor.

One may require repeated two-officer deployments, substantial overtime, supervisory review, extensive paperwork, and downstream court processing. Another may resolve the same condition through targeted deployment, environmental change, coordinated agency action, warning, civil enforcement, or another intervention requiring fewer resources.

If the outcome is comparable, the strategies are not economically equivalent.

This is why officer-hours matter.

Without an estimate of how much time different enforcement activities consume, the City cannot calculate even a basic cost per action, much less a cost per successful outcome. The missing information is particularly consequential when overtime is involved because overtime converts an opportunity cost into an immediately visible additional expenditure.

A mature performance system would therefore report not merely the number of enforcement actions but the personnel resources required to produce them.

That could permit calculations such as cost per criminal summons, cost per legally sufficient criminal summons, cost per resolved quality-of-life complaint, cost per sustained traffic-safety improvement, or cost per successful enforcement outcome.

Those figures would not necessarily determine policy by themselves. A high-cost intervention can be justified when the harm being prevented is sufficiently serious. But the cost would at least become visible.

At present, the public can often see the activity without seeing the price attached to producing it.

Legal Sustainability Should Be Treated as a Performance Measure

The criminal-summons findings also suggest that legal sufficiency should be incorporated into management review.

If a substantial percentage of enforcement actions are rejected because of defects, incorrect legal provisions, illegibility, or other facial deficiencies, that is not merely a court-system statistic. It is information about the quality of the enforcement process.

The Department could track defect and dismissal rates by command, offense category, enforcement initiative, and time period. A sudden increase could identify a training problem, supervisory problem, paperwork problem, or pressure to produce quantity at the expense of accuracy.

The purpose would not be to reward convictions or punish officers because a case was dismissed. Criminal cases fail for many legitimate reasons unrelated to officer performance. The more useful distinction is between substantive case outcomes and defects attributable to the instrument or process itself.

A summons rejected because the wrong copy was transmitted presents a different management issue from a summons dismissed after a contested factual hearing.

The current research already makes that distinction possible. A performance system could institutionalize it.

Distributional Effects Also Belong in the Performance Record

The racial and socioeconomic findings discussed in the preceding section should likewise be integrated into enforcement review rather than treated as a separate annual demographic exercise.

Population disparities alone do not establish why enforcement differences exist. But persistent disparities—particularly when they vary sharply by borough, neighborhood income, or enforcement type—identify questions that require further examination.

A return-based model could compare who bears the enforcement burden with where the underlying public-safety benefit is actually realized.

That is a more useful inquiry than merely publishing race percentages.

If a strategy disproportionately subjects one population to police contact, court obligations, fines, or other burdens, the relevant questions include whether that distribution reflects the condition being targeted, whether similarly situated populations are treated comparably, whether the intervention is producing measurable benefits in the affected communities, and whether those benefits justify the concentration of enforcement.

The same analysis can identify improvements. If a command reduces racial disparity without sacrificing public safety, that is operationally important information. It may identify a strategy that other commands can study and replicate.

A Public Enforcement Return Dashboard

The practical expression of this model would be a public enforcement-return dashboard.

It would not need to publish every operational detail or compromise legitimate law-enforcement interests. Its purpose would be to place the principal inputs and outcomes of major enforcement programs in the same analytical frame.

For each enforcement category, the public record could show activity volume, estimated officer-hours, overtime where attributable, direct and fully loaded cost, legal-sufficiency rates, adjudicative outcomes, repeat enforcement, geographic distribution, racial and socioeconomic distribution, complaint or claim rates where data permits, and the outcome measure corresponding to the stated purpose of the initiative.

The most important feature would not be any individual metric. It would be the relationship among them.

A large increase in enforcement accompanied by a substantial reduction in the targeted harm would look very different from a large increase in enforcement accompanied by high defect rates, repeated activity at the same locations, and no measurable change in the underlying condition.

Both scenarios can currently produce impressive activity statistics.

They should not produce the same assessment of performance.

Count What Matters

The central problem is not that NYPD collects too much data. It is that much of the data describes government action more clearly than government return.

A police department responsible for public safety necessarily exercises discretion about where officers go, what conduct they prioritize, which enforcement mechanism they use, and how much time and money they devote to particular problems. Those decisions involve tradeoffs whether or not the tradeoffs are formally acknowledged.

The accounting system should make them visible.

Enforcement volume can remain part of that system. It is useful information. But volume should be treated as an input or output, not as a substitute for effectiveness.

The more meaningful question is what happened next.

Did the enforcement survive legal review? Did it solve the condition? Did complaints decline? Did serious harm fall? Did the need for repeated intervention decrease? What did the result cost? Could another strategy have produced the same or a better outcome with fewer resources or fewer collateral burdens?

New York City already knows how to count police activity.

The remaining challenge is to measure what that activity actually produces.

About the Author