A heated argument occurs between two lovers and, in a fit of rage, one stabs the other. Or, two drug dealers both want the same piece of turf, and the dispute is settled with automatic weapons. The circumstances that may lead one human being intentionally to kill another are as diverse as the fabric of human relations itself. Yet, even though each individual act of murder is committed within a unique context, if we add them all together we may discern a pattern that sheds light on the question of why murder rates rise and fall over time.

Over the last six years, there has been a dramatic reduction in the number of homicides in the United States. Though still by far the most violent of all Western industrialized nations, from 1993 to 1998 the number of murders in the United States dropped 31 percent, from 24,530 to 16,914, pushing the murder victimization rate down from 9.5 per 100,000 to 6.3. Experts have tried to account for this decline by citing various factors, ranging from the booming economy to shifting demographics, from the decline in the market for crack cocaine to the surging prison population, from more effective policing to tighter gun control legislation. There is, however, no consensus over why the murder rate has fallen so far and so fast.

Immediately before this, from 1985 to 1991, the problem was different, as the murder rate took a sharp turn upward. While scholarly disagreement continues over the cause of this increase, there is one salient and widely acknowledged factor that burst onto the scene in the mid 1980s - crack cocaine. The highly addictive, cheap, and brief duration of a crack high - normally lasting no more than ten minutes - made the inner-city a natural habitat for this drug, where user and seller could remain in close proximity, and the dense population provided a large potential market. Given its profitability, crack selling led to widespread urban violence as rival gangs fought over "turf,” and the homicide rate shot up in those cities where crack was sold. After the murder rate peaked in the early 1990s and then began to fall precipitously, scholars searched for a reason. They are still searching. Several promising theories have been put forward - and legions of politicians have rushed to take credit - but no single theory or policy seems able to account for the magnitude of the recent drop. More than merely undoing the spike caused by crack, the current homicide rate is the lowest since 1967. While it is tempting to think that such a clear turnaround must have its roots in a single causal factor, it is more likely the result of complex interaction among several forces.

ANATOMY OF A MURDER TREND

This is not the first time that there has been such a large reduction in murder in the 20th century. The U.S. murder rate has fluctuated widely over the last one hundred years, reaching a high in 1933 that was close to the most recent peak in 1993. From 1933 to 1938, the murder rate dropped by 30 percent, nearly matching the contemporary decline. But data from this era are sparse and sometimes inaccurate, and experts are unsure what caused the fall. The end of Prohibition in 1933 probably had some effect on stemming the violence that had been associated with the illegal distribution of liquor. But just as significant might have been advances in medical care made during that era, which would have saved many an aggravated assault from becoming a homicide.

Today, we at least know who is committing fewer murders, even if we are not sure why. Based on the research of Carnegie Mellon University criminologist Alfred Blumstein and others, we know that the contemporary decline in homicide is largely the combination of two distinct trends. One is a sharp decrease in homicides committed by youth (those under 25), and the other is a decline in the number of murders involving guns. In a recent paper, "Explaining Recent Trends in U.S. Homicide Rates,” Blumstein and University of Missouri-St. Louis criminologist Richard Rosenfeld have calculated that between 1993 and 1997 the number of homicide arrests for youth fell by 24 percent, which accounts for fully two-thirds of the total decline in the number of homicide arrests during this period. Other data, from the U.S. Bureau of Justice Statistics, show that during the same period the number of gun homicides fell by 28 percent, two-thirds of which was due to youth. Thus, like the increase that preceded it, a large share of the recent decline in homicide is due to the behavior of youth using guns.

But not all of the drop in murder has been due to youth. As Blumstein and Rosenfeld's figures indicate, about a third of the decline in homicide is attributable to those over 25; indeed, the adult homicide rate has been falling since the 1970s, even from 1985 to 1991, while the overall homicide rate was increasing. Another facet of the recent decline is that until lately it has been driven primarily by the largest U.S. cities. In 1995, 40 percent of the national drop in homicide could be accounted for by just six cities. Given its large share of the national population, and its relatively high homicide rate in 1993, New York City's 67 percent reduction in homicide from 1993 to 1998 itself accounts for 17 percent of the national decline during this period. But New York's experience has not been unique; over the same period, the number of homicides has dropped in San Diego by 68 percent, in Boston by 65 percent, in Los Angeles by 60 percent, in San Antonio by 60 percent, in Houston by 43 percent, in New Orleans by 42 percent, in Detroit by 26 percent, in Philadelphia by 23 percent, in Dallas by 21 percent, and in Chicago by 18 percent. Together with New York, these cities account for 8 percent of the national population, but 59 percent of the decline in homicides. For 1999, statistics from the FBI's preliminary Uniform Crime Report indicate that the largest drops are now occurring in smaller cities, such as Nashville, Tennessee, at 50 percent, and Fort Wayne, Indiana, at 41 percent, as the largest urban areas have now bottomed out.

DEMOGRAPHICS

Much of the early speculation on what caused the national drop in murder has centered on demographics. Historically, young people (men especially) have had the highest murder rates; the peak ages for committing murder are between 18 and 24, when the offending rate is nearly triple that of the next oldest age group, 25 to 34. During the 1960s, as the baby-boom generation came of age, the national murder rate rose. Today, as even the youngest of the baby boomers are now 35, their involvement in crime has dropped off sharply.

Since 1990, the number of 18-to-24 year-olds has been declining, and this has led some to suggest that murder is dropping because fewer people are in their prime murder committing years. This has even led to predictions of a coming crime wave, when the larger birth cohort who are now 14 to 17 enter their "high crime” years.

But, this can only be a partial explanation, for it does not account for the fact that the murder offending rate has been falling steadily since 1993, for both 14-to-17 year-olds and 18-to-24 year-olds. Without this rate drop, the dip in the number of 18-to-24 year-olds would have resulted in only a 1 percent national reduction in the number of homicides from 1993 to 1997; with their rate reduction, this age group actually caused a 9 percent drop. And during this same period, the number of 14-to-17 year-olds actually increased, yet the falling murder rate within this birth cohort caused a 6.7 percent reduction in the nation's total homicide count. Together, the falling rates within these two age groups account for well over half of the 25 percent national decline from 1993 to 1997. Clearly, this tells us that the drop in homicide is due not just to the fact that there are fewer young people - who traditionally have been the most likely to commit murder - but also to the fact that fewer young people are now committing murder. We are dealing with a behavioral shift, and not just a demographic one.

One highly controversial variant of the demographic hypothesis has been offered by Steven Levitt, an economist at the University of Chicago, and John Donohue, a law professor at Stanford. They have sought to correlate the recent drop in murder with the legalization of abortion in 1973. In a paper that has received widespread attention in the press, Levitt and Donohue argue that the drop in the number of murders - as well as the declining murder rate - can be explained by the crime-inhibiting effect of planned parenting. The increase in abortion after 1973, they submit, not only reduced the number of 18-to-24 year-olds available to commit murder in the 1990s, but also increased the likelihood that those children who were born were "wanted,” and thus less likely to commit murder when they became adults. But, these researchers are merely speculating based on temporal correlation, and do not offer any specific evidence of a causal link between abortion and murder. They also neglect to account for the decline in the murder rate of adults born before 1973.

LAW AND ORDER

One of the more provocative hypotheses put forward to explain the drop in murder arises out of a quasi-experimental program in policing introduced in New York City. In January 1994, the New York City police department, under its new commissioner William Bratton, instituted a radical policy change. After years of ignoring such "petty” crimes as turnstile hopping, public urination, aggressive panhandling, graffiti, vandalism, and "squeegee operators,” it was decided that such "quality of life” offenses were creating an environment within which more serious crimes could flourish. The NYPD began to crack down on these offenses, in the hope of deterring more serious crime. Support for this policy was self-consciously based on the work of sociologists James Q. Wilson and George Kelling, who argued that "Social psychologists and police officers tend to agree that if a window in a building is broken and is left unrepaired, all the rest of the windows will soon be broken. This is as true in nice neighborhoods as in run-down ones.” Thus, small signs of disorder may lead to more serious offenses.

A demonstration of the "Broken Windows Theory” had been provided many years earlier by Stanford psychologist Philip Zimbardo. In a famous experiment, two automobiles without license plates were parked with their hoods up, one in the Bronx and one in Palo Alto, California. The car in New York started to be stripped within minutes, and was totally destroyed within three days. The car in Palo Alto sat untouched for more than a week. Zimbardo then smashed its windshield with a sledgehammer. Within hours this car too was stripped, turned upside down, and virtually destroyed. Once the appropriate cue had been given that the car was abandoned - relative to the standards of the neighborhood - the results were identical.

New York's recent "experiment” with the Broken Windows Theory has been no less stunning. Since 1993, crime has been cut in half and murder has dropped 67 percent. Proponents of the Broken Windows Theory have claimed credit, noting that many of those arrested for petty crimes were found to have been wanted on more serious criminal charges and/or were illegally carrying firearms. In 1997, for example, a serial murderer was apprehended, whose only previous arrest - resulting in the fingerprinting that led to him - had been for turnstile hopping. Some also argue that, as word spread of the newly aggressive police tactics, many felons began to leave their guns at home, and crime correspondingly decreased. They also claim that many petty criminals were deterred from moving on to more serious offenses after being stopped for smaller ones. Finally, supporters of Broken Windows policies are eager to point out that cities such as Houston and New Orleans copied the methods used in New York, and achieved similar success.

But, the Broken Windows Theory less easily accounts for the drop in murder across the whole nation. Cities such as Los Angeles, where murder dropped 60 percent, have not been known for their innovative police tactics. And, Boston achieved results nearly identical to New York, even though Broken Windows was eschewed in favor of community policing and outreach.

This has led some to argue that Broken Windows policies - and many other local crime initiatives across the country - have simply put more police out on the streets, and that this is what is having an effect on murder rates. While there may be some truth to this, one notes that the number of police in the United States has been increasing modestly since the 1970s, even as the murder rate has fluctuated. And, some have doubted whether the comparatively small 16 percent increase in the number of police per resident (from 1975 to 1998) is enough to have had much of an effect on crime.

Similar problems of timing arise for those who argue that the decline in homicide is explained by an increase in incarceration. There are now more criminals in prison - nearly 2 million - than at any time in American history, and the national incarceration rate has increased by 350 percent since 1972. Yet during this time, the murder rate has varied. Indeed, both the prison population and the rate of incarceration grew steadily from 1985 to 1991, while the murder rate was rising.

IT'S THE ECONOMY, STUPID!

The period from 1993 to 1998 falls in the middle of what, by February 2000, has been the longest sustained period of economic growth in American history. Unemployment and inflation are at thirty-year lows. Even unemployment rates for young men, albeit higher than the national average, have taken a dive. Could the good economy explain the drop in murder?

While murder is not normally committed solely for economic gain, it seems reasonable to think that there may be some link between the economy and murder. Many crimes, after all - from armed robbery to kidnapping to burglary - have an economic motivation. As the economy improves, and people have more opportunities for financial gain, economically motivated crimes might decline, and so too the murders that can accompany them. Also, one might expect that a share of murders are committed out of frustration, anger, or despair, and that these emotions are exacerbated by difficult economic conditions.

As plausible as this theory sounds, however, it does not square with the best available evidence. In four separate studies, employing data ranging from 1933 to 1990, researchers have found no positive correlation between the economy and murder. In one study, Bijou Yang and David Lester, of the Center for the Study of Suicide at Drexel University, found that unemployment was associated with suicide, but not with homicide. In a more comprehensive study, Philip Cook and Gary Zarkin, both economists from Duke University, found that while recessions are indeed positively correlated with robbery and burglary - two obviously economic crimes - the homicide rate was insensitive to business cycle fluctuations. Two further studies, one by Chester Britt of the University of Illinois, and another by David Cantor, from the Bureau of Social Science Research, and Kenneth Land, of the University of Texas, actually found a negative correlation between unemployment and homicide.

While more contemporary research - encompassing the most recent homicide drop - is clearly warranted on this question, recent history provides some anecdotal support to these empirical findings. In the last economic expansion before this one - from 1982 to 1990 - the murder rate did not decline, but rose. And, during the period when New York City's reduction in murder was leading the nation, its unemployment rate was twice as high. Looking farther back, one notes that during the 1930s - the last time the national homicide rate dropped 30 percent - the nation was in the depths of the worst economic depression in its history. So, while the robust U.S. economy might have played some role in the recent decline in murder, there are clearly other forces at work.

CRACK, GUNS, AND YOUTH

It is now time to consider a conclusion that many have resisted that no single factor may easily explain the drop in murder in the 1990s. Neither demographics, nor better policing, nor a high rate of incarceration, nor a growing economy alone will do the job. While each of these factors likely contributed to the overall decline, no one is sufficient to explain it. More likely than any "single bullet” theory is an explanation where the causes are multiple and complex, and the presence of any one factor may facilitate and amplify the effect of others.

One such explanation, which tries to account for the drop in murder in terms of many interacting forces, is that given by Blumstein and Rosenfeld. They speculate that the declining crack market of the early 1990s, combined with improved job opportunities at the low end of the wage market, led to fewer youth choosing drug selling as a profession. This decreased competition for drug markets, they argue, led to a corresponding de-escalation of the arms race amongst dealers that had accompanied the crack explosion in the mid 1980s. Add to this the aggressive police crackdown on guns that was occurring in many cities - from the "stop-and-frisk” policy that was part of the maintenance of public order in New York City to the "Cease Fire” community mediation efforts initiated in Boston - and there was a powerful deterrent to carrying guns. Consequently, as the number of gun confrontations between youth dropped, so did the opportunities for murder. What about the decline in the murder rate for adults? Here Blumstein and Rosenfeld cite the continuing impact of the growth of incarceration, which shows a steady twenty-five-year trend that precisely tracks the drop in the adult homicide rate.

This account has the virtue of explaining both the decline in murder by youth - who drove most of the overall decrease - and also the continuing decline in the adult homicide rate. Drawing together a number of disparate hypotheses, the authors show how the decisions of urban youth, the decline in crack and guns, better policing, the good economy, and the incarceration effect, all might have interacted to produce an outcome that none of these factors alone could have achieved. As the authors put it "Multiple factors are almost certainly responsible for the recent homicide decline, and the effectiveness of any single factor depends on the presence of others.” Of course, such a complex hypothesis is difficult to test directly, and its validity may only be assessed over time, as additional evidence comes in. Yet, it is a clear asset of this theory that it accounts for the decline in murder, at least in part, through a reversal of several of the factors - like the market for crack and the use of guns - that drove the murder increase in the 1980s. It also anticipated the lagging rise and fall of murder in those smaller cities, such as Louisville, Kentucky, where crack markets matured later.

THROUGH A GLASS DARKLY

Are we now confident that we understand the reasons for the recent drop in murder? Can we use this knowledge to sustain the current decrease or to influence murder rates in the future?

Here we face the ultimate problem with any social explanation change. For just as society changed between the 1930s and the 1990s, so too might the environmental determinants responsible for the next increase in murder be different than those that we now face. Understanding homicide is not an exact science; it is always possible that novel phenomena (like the invention of crack cocaine) will arise and alter the causal environment. Given the vicissitudes of human nature, there will inevitably be limits to what we can know.

And, even if we assume that we have some idea of what causes murder, and so possibly can prevent it, how much are we willing to give up to do so? A recent increase in complaints of police brutality in New York City, as the Broken Windows Theory has been implemented, is one place that such questions have been raised. After the accidental police shooting of an unarmed citizen in early 1999, community pressure led to a weakening of police enforcement of "quality-of-life offenses,” resulting in 2,000 fewer citations for petty offenses, and a 16 percent drop in gun arrests, in the first half of 1999. Significantly, this was accompanied by a 6 percent increase in murder. Similar concerns about the number of people in prison, the erosion of our right to bear arms in the face of tighter gun control legislation, and other threats to our civil liberties posed by the fight against crime, raise the question of what price we are willing to pay in order to stop murder.

The prevention of murder is widely agreed to be among the most important of our societal goals. Yet, despite such broad consensus, the tradeoffs and uncertainties that face even the best-intentioned crime control policy may provide natural limits to what we can achieve. But with the stakes so high, even our best efforts seem little enough when trying to understand a topic so weighty as murder.