Explaining recent trends in U.S. homicide rates.

• Author: Blumstein, Alfred

1. THE CHANGING HOMICIDE RATE

   During the past decade some sharp swings have occurred in the homicide rate in the United States. The rate in 1980 was a peak of 10.2 per 100,000 population, and by 1985 it fell to a trough of 7.9. It then climbed a full 24% to a peak of 9.8 in 1991, and has been declining markedly since then, reaching a level of 7.4 in 1996 (see Figure 1) and 6.8 in 1997, which is lower than any annual rate since 1967.(1)

   [Figure 1 ILLUSTRATION OMITTED]

   The jubilation over this decline is mixed with widespread curiosity over the factors that are responsible for it. In this paper, we explore some of those factors, focusing particularly on those whose effects are reasonably measurable and where aggregates may present a misleading picture. In some cases, for example, we identify aspects of some variable contributing to an increase in homicide and other aspects of the same variable contributing to a decrease.

   This is the case, for example, with age. During the late 1980s, young people were contributing to an increase in homicides while older people were committing fewer homicides and contributing to a decrease. In other cases, there are important interactions, for example, between race and age. A large increase in homicide with handguns occurred among young African-Americans in the late 1980s, but we observe no such increase for older African-Americans. In such instances, demographic disaggregation is necessary to isolate the effects being examined. A general theme of our discussion is that it is not productive to think of homicide rates as unitary phenomena. Rather, the recent change in the aggregate homicide rate is the product of several distinct subgroup trends. Any credible explanation-much less forecasting-of the overall change in homicide rates, therefore, must make sense of multiple, interactive, and sometimes countervailing influences.

   Many explanations have been offered for the recent decline in homicide rates. There have been claims, most notably by New York City Mayor Rudolf Giuliani and William Bratton, when he was New York's Police Commissioner, that virtually all of the homicide drop in New York resulted from smart and aggressive policing.(3) Another view attributes the decline to a change in some of the factors that contributed to the growth, most importantly, a reduction in the high rates of firearm homicide committed by (and mostly against) young people, particularly African-Americans.(4) Some of this turnaround may be the result of changes in policing, especially the use of aggressive stop-and-frisk tactics to remove guns from kids, but other factors could well be involved. These could include community efforts to mediate inter-gang disputes, a greater availability of jobs in the booming economy, changing drug markets with diminished roles for young people, and growing incapacitation effects through increases in the prison population. And, looking across the nation, we will find that the effects of changes in the large cities have a dominant effect on the aggregate rates.

   In this article, we assess the influence on homicide trends of some of these changes that have occurred during the last decade. We begin by setting out some basic facts about recent trends in homicide that any credible explanation must confront. These include the differing patterns by age, recent demographic shifts, the role of weapons (particularly handguns) and the domination of the national trends by the changes occurring in the largest cities. We then evaluate several explanations for the drop in homicide, devoting particular attention to explanations which focus on the impact of declining or "maturing" drug markets, growth in incarceration, economic expansion, changes in family structure, and enforcement policy. We conclude with some speculations about the changes in homicide rates likely to occur over the next decade, given certain assumptions about stability or change in the conditions that, in our view, have been most closely associated with past trends.

2. THE RHETORICAL CONTEXT

   We should say at the outset that it is not our intention to engage in explicit forecasting of homicide trends. Forecasting homicide became a popular academic pastime during the rapid increase of the late 1980s and early 1990s. Some of the predictions were accompanied by dire and dramatic warnings of an impending crisis in criminal violence. One academic commentator warned of a coming "blood bath" of juvenile and youth violence that would inevitably accompany the increasing size of the youthful segment of the population, even if their offending rates did not go up.(5) Others characterized a subgroup of active offenders as youthful "super predators" who would terrorize the nation's cities.(6) The motivation for this kind of rhetoric from persons the public looks to for serious guidance on the crime problem is unclear. The effect, however, has been to reinforce an already overheated climate of opinion and policy regarding the homicide problem. We do not wish to contribute further to that rhetorical environment.

   Claims of a rather different sort, coming primarily from public officials, accompanied the drop in homicide rates in one city after another in the early 1990s. As happened in New York, the inclination to account for the local declines in terms of this or that special local initiative apparently was all but irresistible.(7) Not surprisingly, perhaps, very different types of causal rhetoric dominate public discussion during periods of rising and falling crime rates. During periods of increase, explanations tend to emphasize the importance of immutable conditions or forces for which public officials (at least those currently in office) cannot reasonably be held accountable. Popular examples are demographic changes caused by the birth rates of another decade, breakdowns in the family and other key institutions, and a generalized decline in morality and civilized behavior. In contrast, when serious violence is on the decline, the imputed causes are more often located in policies or practices for which public officials are willing or eager to take credit, such as putting more police on the street or more offenders in prison.

   A kind of rhetorical bidding war began to emerge in the early 1990s between some academic criminologists, who tended to explain the decrease in homicide and other violent crimes in terms of factors that were largely beyond the control of policymakers, and policy enthusiasts, who saw the decline as evidence confirming the utility or wisdom of a favored program or practice.(8) This debate has a sterile and dogmatic quality that is unlikely to advance understanding of the multiple and interacting factors responsible for the rise and fall of homicide rates over the past decade. Our view is that policy can make a difference, but the difference it makes is highly dependent on existing levels and trends in violent crime. We elaborate this position later in the paper with reference to the dramatic decline in New York City's homicide rate.(9) Our point here is simply that the debate over whether policy changes are responsible for the drop in homicide has largely overlooked the question of how policy interacts with other factors influencing trends in criminal violence. A key factor with which any policy assessment must contend involves the sharply different trends in homicide by age.

3. CHANGES IN AGE-SPECIFIC HOMICIDE ARREST RATES

   An earlier paper in this Journal presented the striking changes between 1985 and 1992 in age-specific arrest rates for homicide(10) That paper showed that, while the rates for persons age eighteen and younger more than doubled, the rates for those thirty and above declined by about 20-25%.(11) We can now extend those analyses to 1997, and we see some striking changes in the younger group. Figure 2a presents the age-specific arrest rate for murder for the years 1985, which was the last year of a fifteen-year period of very stable age-specific rates, and 1993, which was the peak year of juvenile age-specific rates. We see that, even though the rates for ages twenty and under had more than doubled over this interval, the rates for those over thirty had indeed declined.

   [Figure 2a ILLUSTRATION OMITTED]

   Figure 2b depicts the same 1993 situation along with the figure for 1997, where we see the rates for all ages decline, with the steepest decline around age eighteen, where the growth had been greatest.

   [Figure 2b ILLUSTRATION OMITTED]

   It is instructive to break out these changes in more detail by looking at the time trends for each individual age. Figure 3 depicts the trend for the ages traditionally displaying the peak homicide arrest rates, eighteen through twenty-four. We see how similar those rates were from 1970 through 1985, and then the divergence beginning in about 1985. The rate for the eighteen-year-olds more than doubled by 1991, dropped in 1992, reached a new peak in 1993, and then declined for the next three years. The pattern is similar for the other ages depicted in Figure 3, although the rise in the late 1980s is less steep with increasing age and the decline after 1993 is correspondingly less for the older ages. For youth under eighteen, the pattern is very similar, although the stable base rate in the 1970-1985 period is lower, but in all cases the rate more than doubled by 1993. The pattern for the ages above twenty-four is similarly flat through the mid 1980s, followed by a steady decline for most of the ages.

   [Figure 3 ILLUSTRATION OMITTED]

   These changes for the growth period, 1985 to 1993, and for the decline period, 1993 to 1997, are reflected in Figure 4, which depicts for each age the ratio of the age-specific arrest rate for murder to the rates that prevailed in 1985. Points above the heavy line (at the ratio of one) represent an increase in the rates and points below that line represent a decrease. The upper graph portrays the ratio reached in the peak year...