Murder in New York City

Chapter 1

The Long Sweep and Big Events

Conventional Wisdom versus Reality We rightly think of the United States as a stable country. Even so, the past two and a half centuries have witnessed several major cataclysms and social shifts that should have affected everyone's behavior, including behavior that turns into murder. New York has lived through a revolution and then enemy occupation, many riots, a draining of young men for a highly unpopular civil war, massive immigration from abroad and from the countryside, extraordinary growth, crowding and poverty, the disruption of two foreign wars, and a crushing national depression. The city also suffered several cholera epidemics in the nineteenth century and a lethal influenza epidemic in the twentieth (for which the population data used in this book have been adjusted). These events all make the city an ideal laboratory to test some of the major ideas that link social change to violence. Some of these ideas turn out to be little more than unexamined myths. This chapter challenges several. They include:

1. Cities are cauldrons of murder, so why bother with all this work?
WRONG: For the twentieth century prior to 1958, New York City had lower homicide rates than the United States as a whole. 2. The underlying social forces of mass society cause deviance. Big fosters bad.
WRONG: As New York City grew bigger, it often became safer. Late-twentieth-century violence increased as the city lost population.

3. Crowding leads to deviance and violence, as proved by rat experiments.
WRONG: At times in the late nineteenth and early twentieth centuries, sections of New York City were more crowded than anywhere in the Western world--more crowded even than Bombay, said some--but the city was safer than it is today. People differ from rats in some ways. [note 1]

4. Poverty explains murder.
WRONG: In some of New York City's most miserable periods, murder rates were at their lowest.

5. A corrupt criminal justice system loosens morals and leads to violence.
WRONG: Morals may loosen, but at most identifiable times of immorality and corruption, homicide rates were low.

6. We know what causes violence: young men coming home from war, trained to kill.
WRONG: Postwar times were some of the most peaceful ones we have seen.

7. Riots unleash violence.
WRONG: Riots occur in times that are already violent.

1. Cities are cauldrons of murder. Figure 1.1 shows twentieth-century homicide rates for the whole United States and for New York City. The biggest city in the United States looked benign compared to the nation as a whole--until 1958. Who would have guessed that in the years following, the city's homicide rate would zoom so high? Did New Yorkers even know that a kind of golden era was ending? No wonder that teenage gangs could be seen as romantic and cute, as in West Side Story. From the point of view of the remainder of the twentieth century, the idea that the biggest city could have a lower homicide rate than the country itself is unthinkable. Yet, in the year that West Side Story--a poetic, romantic vision of youth crime and gangs--became a Broadway hit, this was the case. By the late 1980s, the city of West Side Story contributed 10 percent to the total of all homicides for the country. [note 2]

The U.S. homicide rates plotted in figure 1.1 include both city and countryside. Prior to the 1920 census, half the population of the United States lived in places with fewer than 2,500 inhabitants. Strangely, since Frederick Jackson Turner presented his seminal address "The Significance of the Frontier in American History" at the Chicago World's Columbian Exposition in 1893, historians have been much more excited by the census bureau's announcement (based on the 1890 census) of the closing of the frontier than by the population's passing the 50-percent-urban mark. Symbolically the frontier represented opportunity whereas the city, by the twentieth century, represented trouble. That the frontier was the actual site of violence seems irrelevant to popular conceptions--except that liberty and violence have often been linked in theory.

Although city size does bear some relation to violence, at least in the second half of the twentieth century, we must observe that the growth of cities does not inexorably lead to violence. Simply put, if urbanization caused violence, then both sets of rates in figure 1.1 should have moved upward together.

In the United States as a whole, the growth of the urban population as plotted against crime looks like figure 1.2. The smooth lower line is the proportion of the population living in cities, towns, and villages with more than 2,500 inhabitants (the definition of urban used by the census bureau, which is useful in its constancy but hardly reflects most people's sense of the term). The jagged line is the annual homicide rate. (The homicide rate is for the whole United States, not just cities, so it should not be compared to the homicide rate in New York City.) There is a correlation between homicide rate and urbanization, in that they loosely rise and level off together. But probably no one looking at this figure would conclude that cities cause crime--the relationship is much too approximate.

Many other kinds of data corroborate the notion that city size alone does not "cause"--or even have much relation to--violent crime. For example, an appendix to one volume of the 1890 U.S. census lists all homicide arrests for 445 cities of more than 8,000 inhabitants, yet city size and the murder rate barely correlate (r² = .01). Similar information appears for 1903 on 175 cities of more than 25,000 inhabitants (r² = .0025), and in the FBI's Uniform Crime Report for the late twentieth century on the 188 cities of more than 100,000 inhabitants in 1996 (r²= .07). In all of these slight positive relationships between city size and homicide rates there is little prima facie evidence to consider size itself as meaningful. [note 3] Although New York City usually has had the greatest number of murders in the United States, its actual rank in the per capita rates has always been well below most other big cities. As the city grew, its homicide rate varied considerably, with a slight positive relationship between size and murder rate appearing only in the twentieth century. [note 4] Perhaps if all things were held equal then larger city size might be shown to cause more violence, but given this unpromising start, such an effort seems unneeded.

If the commonplace idea that size equals violence is so wrong, where did it originate? The answer is not easy. In the preindustrial era, everyone knew that the countryside was dangerous and that cities were safer. Cities had walls and the gates were closed at night to keep highwaymen and brigands out, not in. In Europe, fortified houses were rural, not urban, architecture. They are charming now, but only because the countryside is so safe.

Why would popular impressions change in the urban industrial era, the nineteenth century? Data were available in the 1890 published census that would have allowed a straightforward graph plotting city size and murder arrest rates. Had someone plotted these data, they would have discouraged the notion that city size and crime were correlated: like the data from the early twentieth century plotted in figure 1.2, the 1890 census data showed no relationship between city size and crime rate. In his great work, Suicide, Emile Durkheim casually pointed to data showing that in 1887 suicide was an urban phenomenon, murder a rural one. [note 5]

As early as 1833, however, in the first American translation of Gustave de Beaumont and Alexis de Tocqueville's book on American penitentiaries, temporary New Yorker Francis Lieber claimed that cities produced more crime than rural regions. Supporting his point with logic (increased opportunity and decreased surveillance) rather than evidence, Lieber advanced arguments that were prescient if not immediately influential. [note 6]

If the idea that size equaled deviance came from theory, it did not come from the theory of the best American urban analyst at the turn of the century, Adna F. Weber. In his massive statistical study, The Growth of Cities, Weber questioned whether the city was more "wicked" than the countryside. Arguing that bars per capita measured morality, he found that in New York State, the smaller cities had more bars than the large ones. Dropping his tongue-in-cheek attitude, Weber argued that the city acted like a "spectroscope" or prism, transforming rural mediocrity into the "highest talent or the lowest criminal." [note 7]

In Europe at about the same time, an influential sociologist, Georg Simmel, was propounding the idea that large cities caused social breakdown. In a brilliant analysis that has permeated twentieth-century social thought, he argued that in large cities people lose connection with one another and social values lose out to rampant individualism, and that mental illness and deviance, including crime, breed the "metropolitan personality." By the time of World War I American sociologists had taken up his ideas. It may be that this is where we get our sense that cities cause crime.

Yet when we look at the beginning of the twentieth century in the United States, using published census data from 1903, we find little to confirm the notion that the growth of cities causes increased crime. Figure 1.3, a simple plot of the populations of cities with more than 25,000 inhabitants against their 1903 murder rates, estimated from arrests, shows, if anything, an inverse relationship between city size and crime. If big cities caused crime, the point cloud should have run from the lower left to the upper right. Clearly, we cannot say that the idea of big cities causing crime came from this evidence plotted in a manner easily available in 1903. A more refined analysis of these same data shows that cities with more than a quarter million inhabitants did tend to have very slightly higher homicide arrest rates as their size increased whereas cities smaller than that had a very slightly negative relationship between size and homicide arrest rates.8 In other words, among big cities considered by themselves, larger population meant slightly higher murder arrest rates; among cities smaller than a quarter million inhabitants, larger population meant lower murder arrest rates.

Finally, to sort out this puzzle, table 1.1 shows the ten largest U.S. cities in 1900 and their 1903 homicide arrest rates (per 100,000). Roger Lane has pointed out that New York City's arrest rates were out of line with those of other American cities, so one must keep in mind that these data are about arrests, not murders. If one looks only at the three largest cities and no further, there seems to be a relationship between size and arrest rates. It may have been as simple as this: The bad data on three cities confirmed a sophisticated European theory and, in the popular imagination, obviated any need to look further. [note 9]

The other possible source of the idea is that there are more total crimes in big cities, even when the crime rate is lower, because there are more people in big cities. [note 10] Perhaps if few thought, or--more appropriately--perceived and felt, in per capita terms, this was the source for the "big cities cause crime" formulation. Most social analysts knew better than to make such a mistake, but if this was popular perception, which then was reinforced by sociological theory, and then, after the mid twentieth century, even came to be true, these confluences may be the source of our misconceptions. By the mid twentieth century, the association of big cites with crime had changed to conform to the theory. Big cities did tend to have more crime per capita in the last third of the twentieth century. Given this circumstance, it was easy to project backward and think that the growth of cities had caused the crime surge.

In essence, the postwar world has reversed centuries-old trends. Violent crime began to increase. Big cities became the sources of problems, not solutions. It is important to remember that these developments are reversals. Our mechanisms for coping and understanding must be centered in this context rather than based on ignorant guesses.

2. Big fosters bad. If the linkage between city size and murder is not a general law, what about another version of the same concept? As a city grows, social pressures increase, but it is the growth, not the absolute size, that causes the problem. In this sociological version of Boyle's law, the city is an urban compression chamber that substitutes population and crime for temperature and pressure. This is a dynamic concept, and thus New York City's history comes into play. Figure 1.4 displays the city's homicide rate and population over the past two centuries. This graph gives a nicely equivocal answer. Although the post-1960s violence burst came with a population decline, in some periods, such as the 1840s and 1850s, population and violence did march upward together. This shows the fragility of basing ideas on time series: because New York grew steadily until the last part of the twentieth century, virtually anything that increased would increase simultaneously with a population increase, whether or not there was any causal relationship between the two. It is only the recent decline in population that belies the rigid correlation between population growth and violence and should make us doubt the pressure-cooker effect. [note 11]

3. Crowding leads to deviance and violence. The portion of the graph in figure 1.4 for the decades just before 1900 also calls into question the linkage of population crowding and crime. The latter part of the nineteenth and first decade of the twentieth centuries were arguably the most crowded times in the city's history, as immigrants flooded in. (The 1898 population burst in figure 1.4 is caused by the city's expansion to include the outer boroughs.) [note 12] These were almost certainly the years of greatest impoverishment, the social welfare system struggling to deal with the millions of newcomers.

4. Poverty explains murder. No doubt most murder victims were poor people, but poverty as a mass phenomenon did not automatically increase homicide rates. The Depression substantiates this: New York City homicides increased for the first two years until 1931 (see the line at 1931 in figure 1.4) and then fell throughout the rest of the Depression, paralleling homicide rates in the United States as a whole. On the one hand, we can say that sudden dearth did not provoke homicides. On the other hand, we cannot recapture deeper cultural influences of poverty. A two-decade depression is not the same as a substrate of continual poverty. [note 13]

A canny analyst here might ask if the Depression effect is not masked by the end of Prohibition two years later, in 1933. Legalization of alcohol should have slowed gangland murders. We cannot easily separate out homicides related to illegal alcohol sales, but we can look solely at women victims on the assumption that most organized-crime-related killings were of men. The numbers show that homicides began to fall before the end of Prohibition. [note 14] The number of male victims peaked in 1931 and then began to fall, whereas the number of female victims began to fall a year later, after 1932. Perhaps the additional fifteen men who died in 1931 represented those who were killed over illegal trade in alcohol. Otherwise, the counts show that all murders fell during the Depression and that this decline simply cannot be due to the end of Prohibition.

5. A corrupt criminal justice system leads to violence. The vertical bars in figure 1.4 also bring into question the notion that corruption in the criminal justice system loosens morals and releases violence. The New York City police were beset with scandals during the post-Civil War era and into the early decades of the twentieth century, a period of low murder rates. The most famous investigation, the Lexow Committee, worked during the mid 1890s, years of the city's lowest rates. None of the corruption scandals included the suppression of homicide statistics, and there has never been a suggestion that the homicide counts were manipulated. [note 15] Police corruption involved vice and its illegal payoffs; murder was not inherently profit-making. In part, the homicide counts were separate from police business: the coroner determined which deaths were homicides and sent the data to the public health department. And until about 1910 the coroner prosecuted homicides. This is not to say that the counts were perfect, but simply to point out that they were never alleged to have been affected by corruption, and thus we can be confident that the period was the safest half century in the city's history.

6. Young men coming home from war cause violence. Just as poverty and crowding are usually called upon to explain crime, so too is war, in particular the return of war veterans. It seems logical: young men, trained to use weapons, return to a society with their neighborhood ties disrupted after being gone for so long. It was a given in U.S. history that a period of violence and disorder caused by these men, north and south, followed the Civil War. [note 16] Perhaps this was true for the whole United States, but there are no national data to support the notion. It was not true for the nation's largest city.

Figure 1.5 shows the city's homicide rates with a vertical line in the year of cessation of every U.S. war. Only in the narrowest interpretation, after two of five wars (World Wars I and II), did homicides increase. More broadly, looking at a few years after each war, we see that only World War I was followed by an increase in homicides. Most significantly, none of the wars caused the burst in violence that has always been imagined.

The Civil War and the Vietnam War show some important similarities: during the conduct of each, there were either high or rising homicide rates. The only similarity between these wars was the widespread contemporary disagreement about the wars themselves, but to link this disagreement with violence seems implausible. The violence of war, we must conclude, occurs in the war itself and not in the homes of the soldiers.

What could have been the origin of the idea that war causes violence? Perhaps the presence of veterans among the convicted, perhaps the knowledge that often felons were allowed to enlist rather than go to prison, or perhaps the presence of wounded men who had to beg to support themselves.

7. Riots unleash violence. Unlike wars, there is a link of sorts between major riots and personal violence. The city's most violent riots occurred in 1806 (Christmas Day Riots), 1834 (Election Riots), 1849 (Astor Place), 1857 (Dead Rabbits Riot), 1863 (Draft Riots), 1870 (Orange Riot), 1935 and 1943 (Harlem Riots), and 1977 (Harlem Blackout Riot). Seven of the major riots occurred during times of high or rising violence rates and only three during eras of low or declining violence. In these three, special circumstances exacerbated the violence, whether ethnic conflict (the 1870 Orange Riots) or severe police repression (Harlem in 1935 and 1943). If it were possible to separate deaths due to rioters from deaths due to police or militias firing upon crowds, it is likely that the riots with crowd-originated violence would be even more closely associated with times of overall high personal violence.

This raises the question of just how different crowd violence is from individual personal violence. Because most riots have a rational explanation, it is assumed that the rioters who do violence are rationally motivated but take their political or protest actions to an unusual level of violence. [note 17] The notion that individuals yield up their personalities to angry mobs and that they become more violent than when alone is a commonplace. But is it not possible that violent individuals simply use the crowd as a mask, becoming more violent in riots?

If one wanted to predict riots, one could start by considering the general level of personal violence in a particular society. The possible paths to these results are multiple: violence creates more angry people carrying weapons for defense, violence becomes a more acceptable solution to problems, violent individuals gain power, public violence begets imitation. The general level of violence in a society does not trigger riots: they begin for reasons. But a more violent culture produces much greater violence during rioting than a less violent culture does.

High Violence, Low Violence: When and Why We can define periods of highs and lows in many ways. We can discuss what is high for New York City, or high for the whole United States, or high for a particular time period, or high compared to other similar places, say Liverpool or London. Figure 1.6 initiates this discussion with New York City's rates, smoothed, with peak years marked by vertical lines. Two related features of this picture can be visualized as a horizontal band of "normal" New York City homicide rates that covers the period 1800 to about 1960, and a triptychlike verticalsectioning of the graph into three eras or waves.

The normal range of homicide rates for the city has fluctuated between about three and six per hundred thousand for most of the nineteenth and twentieth centuries. Even alone, these rates represent dramatic variation, a tripling from crest to trough. But a rogue tidal wave of violence in the last quarter of the twentieth century, which appears to be receding as abruptly as did the earlier smaller crests, has dwarfed these earlier swings. The regularity of these patterns confirms the unique magnitude of the late-twentieth-century violence boom.

Including both the normal range and the tidal wave, there have been three distinct periods of about sixty years each. The first two waves crested about 1864 and 1931, and the rogue wave crested in 1991. The troughs occurred in the 1820s-1830s, around 1890-1900, and in the late 1940s-early 1950s.18 The periodicity of these waves, though regular, may not be meaningful because, as the dozens of individual cases discussed in this book make clear, each homicide has enough chance in it to make any grand wave-based theory suspect. The regular periodicity of the peaks raises two questions: Was there a peak around 1740? A trough in 1770? And will there be another peak in 2050? The data for the eighteenth century are simply too scattered to test this graphic implication.

These waves of violence suggest that there are distinct points at which violence changes course. The downturns are much more abrupt than the troughs, the times when low violence rates slowly and erratically turn upward again. When the limit to social tolerance is reached, for a whole range of reasons, violence abruptly diminishes. Then, when some lower level of violence has been achieved, the mechanisms for control and the value of peace get forgotten, and a slow rebirth of violence begins. Though it is vague in operational mechanisms--"violence is in the air"--the idea that eras encourage or repress violence across society is supported by evidence. White and black, male and female, very different kinds of homicide rates parallel one another through these waves of change. There is no reason that this should be--given gender differences and demographic and economic racial differences in murder--unless violence is in the air, a contagious virus. [note 19]

The three periods of sustained increases in violence are surprisingly similar in duration, about forty years each, though the increase from 1950 to 1990 looks longer because the slope is so much steeper. We can only speculate about why this has happened; the wisest thing is to stay with the empirical observation. Identifying historical cycles seems to be an unavoidable adjunct to looking at sweeps of one hundred years or more. Although the exercise is fascinating, it presents the problem that we like to believe that human events are driven by more human forces than the kind of cyclical forces driving climate, planetary orbits, and other nonhuman events.

What do we do with the cycles, once identified? In reintroducing his father's prescient analysis of American political cycles, Arthur Schlesinger Jr. never committed himself to a cycle theory of history, even as "legatee" of one of the most subtle of such analysts. As he concluded, the determinism implied in any cyclical theory "violates our deepest human instincts" such as freedom and responsibility. [note 20] The Schlesingers, father and son, handled this problem for political cycles by arguing that there are eras of optimism and reform, followed by conservatism and retreat.

One might extend such thinking to violence control, arguing that rising violence provokes a multitude of control efforts, many of which have long lags before their effects show up in the murder rates, and that the cumulative effect ultimately drives the rates down. When the murder rate ebbs, control efforts get relaxed, thus creating the multiple conditions causing the next upswing.

One can review the terrible period of increasing violence that began in the late 1950s as just such an era of slowly cumulating efforts--efforts that seemed not to work, at least initially. Everything from child nurturing (Head Start, for example) to increasing penal measures ("heads off," so to speak) may well have cumulated by 1991 to reverse what had seemed an irreversible tide. In past cycles, after three decades of decreasing violence, the violence reduction efforts fade from the policy agenda, laying the groundwork for the next upswing and a forced reinvention of prevention policies.

This gloomy scenario conforms with the position of many policy analysts in other fields, who describe "issue-attention cycles," as Anthony Downs identified them. [note 21] The primary insight that serious social issues fall into and out of the policy spotlight, irrespective of the actual social problems, supports the notion that violence cycles may not be natural at all, but influenced by policy vacillations. Confirmation of this would require measuring social effort: from the family to the schools to churches to popular culture to a myriad of virtually unmeasurable yet important activities.

Since the Civil War if not earlier, social observers have attributed violence at home to returning warriors. Turn this analysis of war and violence a different way and observe that periods of relatively low violence follow wars. Three, perhaps four, of the long periods of decline in violence follow wars--the Revolution (speculation, given lack of data), the War of 1812, the Civil War, and World War II. Another followed the most traumatic economic depression the country has experienced. The exception--as it is to everything--is the decline of the 1990s, unless one counts the Cold War. Perhaps we should, except that unlike the earlier wars, the Cold War did not have soldiers returning from the trauma of violence.

The one exception and explanatory problem comes from the war in Vietnam, which should have caused a decline in the 1970s. Perhaps it would have done so except that the rogue wave was already crashing ashore. This generalization-- peace at home following war--holds if we lump the Vietnam and Cold Wars together, or if we rephrase it as "peacetime brings social peace." Both of these locutions analytically differ from that which links war to violence. For one thing, the notion that war begets violence is usually linked to the idea that military mobilization releases the beast within young men, and that this beast cannot be recaptured quickly. Analogizing to police officers unable to control their aggression after a dangerous pursuit, their adrenaline pumping, who attack the offender, the idea is that soldiers return to civilian life with their aggression and adrenaline pumping, ready to fight. This sounds believable until we consider it a little more precisely: soldiers are carefully trained to be aggressive in the right direction and to control their natural impulses, which typically include flight, dissociation, and withdrawal. A high proportion of soldiers do not experience battle. The "pumped" police officers are not pumped three months later. The soldiers return to civil life anxious to escape their military life. The post-Civil War phenomenon of solders flooding into prisons was simply an artifact of the widespread mobilization: it would have been hard to find any young male who had not had a war experience. The linkage of war and violence combines bad reasoning and little research.

In contrast, the linkage of peace and low violence is a more interesting proposition. What is special about peace? Not every peace lacks a standing army, so it is not actual military training. On the other hand, peace does mean prosperity in a very specific sense: young men follow occupational tracks, even if menial ones. And occupations help individuals develop a future orientation, increase the likelihood of family formation, and make late-night hours and drinking all but impossible.

Capturing Cold War tensions, West Side Story subtly tells us that gang warfare over turf is as old as conflict in the fifteenth century and as foolish as the carefully orchestrated turf violence in Europe during the Cold War. The musical's parallel between the fifteenth century and the 1950s has some historical accuracy: the upper classes in the early modern urban world had access to weapons and engaged in violence against rival groups. High society was violent then, and only the wealthy could afford weapons of quality. Access to weapons has drifted downward, as have the codes that govern the behavior of an armed class. Called the "civilizing process" by Norbert Elias, rules governing interpersonal violence began in late medieval courtly codes (hence the word courtesy).

The explicit conflict in the musical is over turf and ethnicity--Puerto Ricans versus "whites" who, we are reminded carefully, are themselves second-generation immigrants. The police, represented by Officer Krupke, are careerist and racially prejudiced; Officer Krupke is part of the problem in driving apart racial groups. There is no hope from intellectuals, either: a sociological analysis of juvenile delinquency as a "social disease" is mocked. The only saving virtue is young love, which obliterates racial barriers.

Two exchanges of the gang members with Doc, the owner of the candy store hangout, make these themes clear. He chides them for the escalating violence, "You kids'll make this world lousy," and they retort, "We didn't make it, Doc." Later he asks them, "Why do you kids live like there's a war on?" Here the underlying causes are shifted to the adult world and paralleled with the geopolitical fears of the Cold War era.

The early decades of the twenty-first century may have low homicide rates and bring social peace and boredom. The relaxation of the social effort to preserve peace will ultimately lead to rising violence. Then, on the upswing, there will be riots, new suspicion of big cities, and hand-wringing over the American character. The challenge for the next two decades is to maintain the social effort to preserve peace, to try to match other Western nations, and to consider every homicide deterred a major success.

---

Chapter 1 Notes

1. Horace E. Flack, "Congestion in Cities," in Cyclopedia of American Government, ed. Andrew C. McGlaughlin and Albert Bushnell Hart (New York: Appleton, 1914), 1:380-381, citing Lawrence Veiller.
2. Alfred Blumstein and Richard Rosenfeld, "Explaining Recent Trends in U.S. Homicide Rates," Journal of Criminal Law and Criminology 88 (1998): 1175-1216.
3. Arrests are a better guide to actual murders than trials, convictions, or even the nineteenth-century census vital statistics, but none are as good as coroners' reports or local vital statistics. Sources: U.S. Census Office, Report on the Social Statistics of Cities in the United States at the Eleventh Census: 1890 (Washington, D.C.:, Government Printing Office [hereafter GPO], 1895); U.S. Bureau of the Census, Statistics of Cities Having a Population of over 25,000, 1902 and 1903, bulletin 20 (Washington, D.C.: GPO, 1905); U.S. Department of the Interior, Census Office, Report on Crime, Pauperism, and Benevolence in the United States at the Eleventh Census, 1890, part 1, Analysis (Washington, D.C.: GPO, 1895-1896). For 1996-1997, FBI Uniform Crime Reports (Nov. 23, 1997) (U.S. Department of Justice [hereafter USDOJ], FBI, Washington, D.C. 20535), internet file on cities of more than 100,000 inhabitants.
4. Compared to other cities with populations greater than 25,000, in 1903 it was twenty-first, in 1890 tenth, and in 1996 sixty-seventh. Of course, there were more cities with reported murder data in the population greater than 100,000 group in 1996 (188) than in 1890 (28), which makes the comparisons problematic. Correlation of New York's size and homicide rate in the twentieth century: r² = .09.
5. Emile Durkheim, Suicide: A Study in Sociology, trans. John A. Spaulding and George Simpson (London: Routledge & Kegan Paul, 1952), 353. Although Suicide appeared in French in 1897, it did not appear in English until 1951. The lack of an English translation until the mid twentieth century plus the lack of emphasis in Durkheim's throwaway comment on homicide may explain why this empirical critique of the correlation between urbanization and violence has had so little impact.
6. G. de Beaumont and A. de Tocqueville, On the Penitentiary System in the United States and Its Application in France: With an Appendix on Penal Colonies, and Also, Statistical Notes, trans. Francis Lieber (Philadelphia: Carey, Lea & Blanchard, 1833), xx-xxii.
7. Adna Ferrin Weber, The Growth of Cities in the Nineteenth Century: A Study in Statistics (New York: published for Columbia University by the Macmillan Co., 1899), 407, 442.
8. Big city correlation: r = .2; smaller cities: r = -.14.
9. Roger Lane, "The Social Meaning of Homicide Trends in America," in Violence in America, vol. 1, The History of Crime, ed. Ted Robert Gurr (Newbury Park, Calif.: Sage, 1989), 55-79, 65.
10. Noted by Roger Lane in a pioneering article, "Crime and Criminal Statistics in Nineteenth-Century Boston," Journal of Social History 2 (1968): 157-163.
11. Weber, Growth, 459-468, contrasts New York City's denser wards with sections of other Western cities, notably with London's Bethnal Green, but confidently shows how the density decreases as an area ages, calling the whole process "city-building." Like most progressives of his time, Weber saw in the rise of suburbs the hope for alleviating crowding (475).
12. See Kenneth T. Jackson, "100 Years of Being Really Big," New York Times, Dec. 28, 1997.
13. The notion of an underclass has such historical ambiguity that it is almost impossible to employ in an empirical analysis. For essays on the historical context in the United States, see Michael B. Katz, ed., The "Underclass" Debate: Views from History (Princeton, N.J.: Princeton University Press, 1993). The lack of a relationship between unemployment and many kinds of crime probably comes as a surprise to many people: see James Q. Wilson and Philip J. Cook, "Unemployment and Crime--What Is the Connection?" The Public Interest (spring 1985): 3-8; Philip J. Cook and Gary A. Zarkin, "Crime and the Business Cycle," Journal of Legal Studies 14 (Jan. 1985): 115-128.
14. Counts of murder victims by gender allow us to estimate whether the end of Prohibition lowered murder rates:
    Year---Women---Men
    1928---66------336
    1929---91------334
    1930---86------408
    1931---113-----475
    1932---119-----460
    1933---103-----438
    1934---92------366
    Note that even in Chicago, the estimates of murders related to the "Beer Wars" and other gang killings to control the illegal liquor trade amounted at most to 10.7 percent of all murders, 1923-1926.
    Based on the 2,001 total Chicago murders in Chicago Board of Health, Report of the Department of Health, vital statistics (Chicago, 1932), p. 1138; and 215 gang murders for all of Cook County reported by Illinois Association for Criminal Justice, Organized Crime in Chicago, by John Landesco, with a new introduction by Mark H. Haller (Chicago: University of Chicago Press, 1968), p. 97.
15. New York (State) Legislature, Senate Committee on Police Dept. of the City of New York, Report of Special Committee Appointed to Investigate the Police Department of the City of New York, Transmitted to the Legislature January 18, 1895 [Lexow Committee] (Albany, N.Y.: J. B. Lyon, state printer, 1895).
16. Edith Abbott, "The Civil War and the Crime Wave of 1865-70," Social Service Review (June 1927): 212-234. I must point out that I earlier presented evidence that strongly contradicted this theory, but ignored it because the idea was so compelling: Eric H. Monkkonen, Police in Urban America, 1860-1920 (New York: Cambridge University Press, 1981), 77 and 81.
17. The literature on this topic is enormous. For historians, the starting point is George Rudé, The Crowd in History: A Study of Popular Disturbances in France and England, 1730-1848 (New York: Wiley, 1964).
18. The turn-of-the-century trough shown here parallels one previously identified by comparing murder arrest data for the largest U.S. cities, 1860-1920, which showed a bottom reversal point in 1893. See Eric H. Monkkonen and Catrien Bijleveld, "Cross-Sectional and Dynamic Analysis of the Concomitants of Police Behavior," Historical Methods (spring 1991): 16-24. These data graph a decline from the Civil War and an increase from 1893 until 1920. The sixty-year series was not long enough to capture the repeated cycles of homicide.
19. This is not an entirely crazy idea. In 1998 George Edward Tita chaired an American Society of Criminology panel, "In Search of Diffusion: Methods for Identifying Changes in the Spatial and Temporal Distribution of Homicides," which discussed the spread of homicide and violence from one neighborhood to another.
20. Arthur M. Schlesinger Jr., The Cycles of American History (Boston: Houghton Mifflin, 1986), 25, 420. Plotting the homicide rates against Schlesinger's cycles or against those proposed by David M. Gordon, Richard Edwards, and Michael Reich in Segmented Work, Divided Workers: The Historical Transformation of Labor in the United States (New York: Cambridge University Press, 1982), 9, has virtually no predictive power.
21. The literature on the issue-attention cycle is extensive, from Anthony Downs, "The 'Issue-Attention Cycle,'" The Public Interest 28 (1972): 38-50, to Frank R. Baumgartner and Bryan D. Jones, Agendas and Instability in American Politics (Chicago: University of Chicago Press, 1993).