Crime Geography and the Lived Safety Map, 1989–2001

The Peak

The city New Yorkers inhabited in 1989 and 1990 was not the crime-ridden city of tabloid rhetoric — it was the crime-ridden city of the daily commute, which is a different thing. The difference is the difference between an abstraction and a behavior. The number — 2,245 murders in 1990, the highest count in the city’s modern recorded history, against 1,905 in 1989 and 2,154 in 1991 — was the aggregate that the police tracked on paper.¹² What the people who produced that number by being present in the city on those days produced in their own lives was something smaller and more practical: a set of rules.

The conditions that produced the number had been assembling for years. The crack cocaine market reached full commercial amplitude in New York around 1985 to 1988 and was still at or near peak intensity in 1989. The open-air markets ran in Washington Heights, in Hunts Point and Mott Haven in the South Bronx, in Bushwick and Brownsville in Brooklyn, in East Harlem and central Harlem. The market’s primary economic activity was the sale of small quantities of crack at street corners by crews who worked in shifts; its primary violence was the violence of market enforcement — turf boundaries, debt collection, retaliation. The connection between the drug market and the homicide rate in the peak years was not incidental: by the most careful assessments of the period, the crack lifecycle was the single most important structural factor driving the spike.³

Against this the city maintained two separate April and August markers for the year 1989 that had concentrated public attention before it turned to the statistics. On April 19, a woman jogging near the 102nd Street transverse in Central Park was assaulted and left for dead; five Black and Latino teenagers were indicted in connection with the attack.⁴ On August 23, Yusuf Hawkins, sixteen, was shot and killed by a white mob in Bensonhurst — he had come to the neighborhood to look at a used car and was mistaken for someone else entirely.⁵ The marches through Bensonhurst that followed ran into the fall. The two cases occupied adjacent columns in the city’s newspapers for months, and together they supplied the emotional charge to the 1989 mayoral election and to the policing climate that followed it.

What ordinary New Yorkers did with the aggregate was internalize a rule-set that functioned as a kind of distributed safety protocol, never written down, transmitted through conversation and observation and occasional direct experience. Parks were avoided after dark; Central Park’s south end, below 72nd Street, was traversable by day, but the north end, above 100th, was understood to become a different environment by roughly four in the afternoon. ATMs were not used after nine in the evening — the mirrors and cameras at the slot did not yet exist (the ATM Safety Act mandating them was signed in 1996), and the vestibule ATM was understood as a trap.⁶ Cash was counted inside the bank branch before it was put away; a Walkman cord kept visible above a coat collar was a signal advertising something worth taking; certain sneakers — Air Jordans, particular Nikes — were understood by teenagers as robbery targets. After dark, the rule was to walk in the street rather than past doorways. The rule was not theoretical. People who had not had occasion to apply it yet knew it anyway, transmitted through the ordinary channels of a city telling itself how to survive.

The subway ran a parallel system of risk. In 1990, the transit system recorded roughly 17,500 felonies — nearly forty-eight a day, system-wide.⁷ Riders held to the lit section of the platform rather than the ends; they took the first car, nearest the motorman, or the last, nearest the conductor; eye contact was widely understood as provocation and widely avoided. Women on the transit system navigated an additional layer that was neither named nor acknowledged — groping, public masturbation, the press of men on crowded cars — for which there was no common vocabulary, no transit-police response category of any practical consequence, and no reporting mechanism; it was treated, by the system and by most of the people inside it, as a condition of the commute rather than as an offense against it. That subject belongs to another chapter.

The rule-set was not paranoia. It was the applied arithmetic of a city that recorded, in its worst year, roughly six killings a day and several hundred felonies more. The two-wallet system was what the arithmetic looked like in a coat pocket.

The Precincts

The citywide aggregate of 2,245 murders in 1990 was the sum of local arithmetics that varied so wildly from one precinct to another as to make the aggregate nearly meaningless as a description of any particular person’s experience.¹ Some outer-borough and upper-Manhattan precincts absorbed scores of homicides in a single year; several precincts serving Midtown, the Upper East Side, and parts of the Upper West Side west of Amsterdam went entire calendar years without a single murder. The city that had 2,245 killings was also a city in which many of its residents had no direct daily encounter with lethal violence — not because the killing was not happening, but because it was happening somewhere they did not live and rarely went.

Where it was happening was not random. The geography of homicide in the early 1990s was the geography of the open-air drug market, which was itself the geography of the urban disinvestment that had concentrated poverty and stripped institutional capacity from certain neighborhoods across the 1970s and 1980s. East New York and Brownsville in Brooklyn; Hunts Point, Mott Haven, and Morrisania in the South Bronx; Washington Heights in upper Manhattan during the crack market’s peak years; central Harlem on certain corridors — these were the places where the number was not a statistic but a proximity, heard and sometimes seen. The 1970s arson wave had destroyed much of the housing stock in some South Bronx census tracts; what remained was the physical inheritance within which the drug market operated in 1989.³ The relationship was causal in both directions: disinvestment had created the conditions, and the violence further impeded the reinvestment that might have altered them.

Where it was not happening — or not at the same scale — was the Upper East Side above 60th Street, the blocks west of Amsterdam on the Upper West Side, the West Village’s brownstone interiors, the interior of Park Slope, the established residential blocks of Brooklyn Heights. The architecture of those neighborhoods was part of the protection: doormen buildings whose lobbies were attended around the clock; limestone and brownstone facades that communicated, to anyone approaching them, the presence of resources whose owners had organized to defend them; locked vestibules that interrupted the flow between street and resident. The aggregate homicide rate for the city included the Upper East Side, but not in any way that felt present to the people who lived there.

The victims of the peak-years violence were, by the measure of published analyses of the homicide data, overwhelmingly Black and Latino — studies of the 1990 to 1991 period found that roughly seven-eighths of murder victims fell into these categories.¹ This was not a consequence of race in any biological sense; it was a consequence of geography, and the geography was the consequence of decades of policy. The neighborhoods that had been redlined, that had lost manufacturing employment as it departed the Northeast in the 1970s, that had been left without adequate sanitation and building inspection during the fiscal crisis, that had lost population to arson and abandonment — those neighborhoods were the neighborhoods of the crack market, and the crack market’s violence fell on the people who lived there.

The official map was a single citywide number. The lived map was a 24-hour cognitive load — lighter where doormen attended lobbies and car services waited at curbs, heavier where the streets required a routing calculation on every walk to the bus stop and a different calculation at night. Both maps were accurate descriptions of the same city. They described different cities for the people inside them.

The Turn

The decline in citywide crime through the 1990s did not begin on January 1, 1994, when Rudolph Giuliani was inaugurated as mayor. It began, by the official numbers, in 1991. The citywide homicide count fell from 2,245 in 1990 to 2,154 in 1991, to roughly 1,995 in 1992, to 1,927 in 1993 — four consecutive years of decline before William Bratton moved from the Transit Police to the NYPD commissioner’s office.¹ The conventional narrative attributing the turn to Giuliani and Bratton elides the administration under which the decline started: David Dinkins, who had run on public safety and whose Safe Streets, Safe City program — funded by a city and state income-tax surcharge enacted in 1990 — had been adding officers to street patrol since 1991. The pin maps and the precinct commanders are one part of the story. The other part is the officers who were already there when they arrived.

Bratton’s earlier contribution was on the transit system, where he had taken command in 1990 under Dinkins. The aggressive prosecution of fare evasion beginning in 1990 and 1991 — turnstile-jumping as the enforcement entry point, under the theoretical framework that visible disorder invited more serious disorder — produced a measurable result. Transit felonies fell from roughly 17,500 in 1990 to around 4,000 by 1997.⁷⁸ The graffiti-free fleet had been secured as of May 12, 1989, when the last tagged car was withdrawn from passenger service; by the early 1990s, the rolling stock ran in clean stainless steel while many station walls remained streaked. The visual contrast — clean trains, dirty stations — was one of the period’s characteristic incongruities. Whether the fare-evasion campaign itself caused the felony decline, or whether something else was driving it, was not settled by the results. The academic literature on the question remains split.⁹

Giuliani’s January 1, 1994 inauguration brought Bratton to the police commissioner’s office. The CompStat system — a weekly meeting at which precinct commanders were required to account for crime trends in their areas against the previous week, month, and year, with resource deployment adjusted in real time to respond to the pin maps — was introduced in 1994 by Bratton and his deputy Jack Maple.⁸ The method was genuinely new as a management technology. Whether it caused, accelerated, or simply tracked a decline whose primary engine lay elsewhere was already contested among criminologists as the numbers fell.

The cost side of the enforcement turn appeared almost simultaneously with the statistical improvement. In December 1994, Anthony Baez, twenty-nine, died of asphyxiation after Officer Francis Livoti applied a chokehold at the family’s house on Cameron Place in the Bronx — the initial contact had been over a football that struck a parked police car.¹⁰ Livoti was acquitted at state trial and later convicted in federal court on a civil-rights charge. The case was not aberrant in the context of the period; it was illustrative of something the Mollen Commission had been documenting since 1992.

The Commission had been established by executive order of Mayor Dinkins in July 1992, chaired by retired Appellate Division judge Milton Mollen, in the immediate wake of the May 1992 arrest of Detective Michael Dowd of the 75th Precinct in East New York — a veteran officer who had been stealing drugs from dealers and selling them, at times while in uniform, for years. The Internal Affairs Bureau had received complaints about Dowd as early as 1986 and had not acted on them. The Commission’s report, released July 7, 1994, documented that this was not an individual failure.¹¹ Corruption of the kind Dowd had practiced was concentrated in roughly ten precincts — all of them high-crime, all of them serving predominantly Black and Latino neighborhoods — and the primary corruption type was drug-related: officers stealing narcotics and cash from dealers, selling the narcotics back into the market, in some cases shooting the dealers they robbed. The anti-corruption apparatus had exhibited, in the Commission’s characterization, “willful blindness” at multiple levels of command.

The case that brought the Commission’s findings into public relief in the starkest possible terms was in the 30th Precinct, in central Harlem. Sergeant Kevin P. Nannery led a unit his subordinates called “Nannery’s Raiders,” which had operated a drug-theft and resale ring from inside the precinct house itself through the early 1990s — making false radio calls to cover illegal searches of dealers’ apartments, taking the drugs and cash, selling stolen narcotics at half-market price. On September 28, 1994, a coordinated federal and NYPD operation arrested twenty-nine officers of the precinct in a single day; the eventual total reached thirty-three.¹² More than a hundred state criminal cases were subsequently dismissed, the convictions having rested on officer testimony now known to have been fabricated. For residents of central Harlem who had been arrested, charged, or convicted on the testimony of officers who had been running the drug trade, the disclosure was not a revelation about bad actors. It was confirmation of something they had experienced as ordinary fact.

The Decline

The numbers fell in a way that was, by any standard, dramatic. From 2,245 murders in 1990, the count reached 1,582 in 1994, fell below 1,000 by the mid-decade figures in the most-cited series, and arrived at 633 in 1998 — a level not recorded since the late 1950s.¹² By 2001, the number stood at 671. Robbery fell from roughly 100,000 incidents in 1990 to under 28,000 by 2001. Burglary, felony assault, and auto theft each fell by two-thirds or more across the decade. The scale of the change was without precedent in the city’s recorded history, and it occurred faster and earlier than in any comparable American city.

What the change felt like on the ground was not what a chart communicates. In the South Bronx and central Harlem in 1989, sirens — police, ambulance, fire — had been near-constant; a quiet half-hour was a disruption of the ambient noise, not the norm. By the late 1990s a quiet half-hour on a residential block in Morris Heights was ordinary. The corner crews that had operated openly on certain intersections through the early 1990s had largely dispersed; the open-air drug market had contracted, moved indoors, or simply shrunk. The car-alarm cycle — the six-tone Viper and Hornet patterns that had run continuously through the night in high-theft areas — sounded less often because fewer cars were being broken into. The empty lots that had stood on certain Bronx blocks since the arson wave of the 1970s were slowly filling with rebuilt walkups and low-rise infill housing. The decline was not primarily a statistical artifact. It was a sensory fact in the places where the peak years had been worst.¹

Beneath the numbers, the cases accumulated. On August 9, 1997, Abner Louima, a Haitian immigrant, was arrested outside a Brooklyn nightclub following a street fight; in the bathroom of the 70th Precinct station house, Officer Justin Volpe sodomized him with a wooden stick.¹³ Volpe pleaded guilty in 1999; he was later sentenced to thirty years. Three other officers were convicted on related charges. The case was prosecuted, the convictions obtained, the sentences served. It did not produce a commission.

On February 4, 1999, at approximately 12:40 in the morning, Amadou Diallo — twenty-three years old, a street vendor originally from Guinea, living at 1157 Wheeler Avenue in Soundview in the Bronx — was shot in the vestibule of his building by four plainclothes officers of the Street Crime Unit: Sean Carroll, Edward McMellon, Kenneth Boss, and Richard Murphy. Forty-one rounds were fired; nineteen struck him. He had reached for his wallet.¹⁴ Civil-disobedience protests at One Police Plaza in Lower Manhattan ran through March and April 1999; well over a thousand people were arrested in the course of the demonstrations. The four officers were tried in Albany — the case having been moved out of the Bronx on a change-of-venue application — and acquitted on all charges on February 25, 2000.¹⁴

On December 1, 1999, New York Attorney General Eliot Spitzer released the first systematic public analysis of the NYPD’s stop-and-frisk practices, based on 1998 data.¹⁵ The report documented that Black New Yorkers were stopped by police at a rate disproportionate to their share of the population and to their share of arrests for violent crime; Latino New Yorkers were stopped at similarly disproportionate rates. The Street Crime Unit — the unit whose officers had shot Amadou Diallo — stopped Black men at rates several times their share of the city’s male population. The report was not an allegation; it was an accounting of department records.

On March 16, 2000, Patrick Dorismond, a security guard and the son of a prominent Haitian musician, was shot and killed by an undercover narcotics officer on Eighth Avenue and 38th Street.¹⁶ The encounter had begun when the officer, working an undercover sting, approached Dorismond and asked him to help obtain drugs; Dorismond declined and a confrontation followed. Mayor Giuliani’s decision to release Dorismond’s sealed juvenile record in the days after his death generated a second extended wave of protest.

The decline was real and it was felt, in the neighborhoods where the worst years had concentrated, as a genuine physical lightening. The methods that produced the statistical change were applied, by the Spitzer report’s own accounting, disproportionately to the residents of those same neighborhoods. Both facts were simultaneously true. Neither canceled the other.

What the Map Was

Why the numbers fell is not agreed upon. The framework most consistently supported by the data centers on the crack cocaine lifecycle: the NYC crack market peaked between roughly 1985 and 1988 and was already contracting by the early 1990s, as the generation of users aged out and fewer new initiates entered; the open-air market violence that had driven the homicide spike in turn declined as the market consolidated and contracted. This framework, developed by researchers including Johnson, Golub, and Dunlap, has the advantage of accounting for the timing — the decline began before the Giuliani-Bratton policies were instituted — and for the geography: cities with very different policing strategies experienced similar declines over the same period.³ Bratton and Kelling have argued, with the support of the transit system’s own data, that targeted enforcement of visible disorder drove the results, beginning in the subways and extending to the streets from 1994.⁸ Levitt’s multivariate analysis assigns weight to several factors simultaneously — the decline of the crack market, increased police hiring levels, rising incarceration rates, and, controversially, the long-lag effect of the Roe v. Wade decision on the cohort that reached peak crime-commission age in the early 1990s — and assigns CompStat a modest marginal contribution rather than a decisive one.⁹ The dispute is genuine and has not resolved in the peer-reviewed literature.

The geographic asymmetry of the enforcement costs is not in dispute. The stop-and-frisk apparatus documented in the Spitzer report was concentrated in the same neighborhoods that had absorbed the worst of the peak-years violence; the Black and Latino men who were stopped, spread against walls, and sent on their way without arrest were disproportionately residents of the precincts where the murder counts had been highest and had fallen furthest.¹⁵ The statistical safety that the aggregate number expressed was purchased in part through that surveillance. The purchase and the cost fell on the same geography.

Beneath the headline crime numbers, the legal armature was also shifting. The Violence Against Women Act of September 13, 1994 provided the first federal acknowledgment that domestic violence and sexual assault were federal crimes requiring federal resources; New York’s mandatory-arrest provisions for family offenses took effect in 1995, replacing the previous framework under which officers had typically “separated the combatants” rather than made an arrest. These changes did not generate the same public attention as the homicide counts, but they registered in the daily rule-sets — particularly for women, for whom the risk map had always included categories of danger the official crime statistics counted imperfectly or not at all.

By 2001, the physical city had changed enough that a resident of central Harlem or the South Bronx in 1989 returning to those blocks would have found them, by the ordinary measures of street safety, substantially different. Whether the change was durable, whether its methods could have been deployed with less concentrated cost, and who ultimately bore the burden of producing it — those questions the numbers did not answer.