Declining homicide in New York City: a tale of two trends.

• Author: Fagan, Jeffrey

The mass media pay plenty of attention to crime and violence in the United States, but very few of the big stories on the American crime beat can be classified as good news. The drive-by shootings and carjackings that illuminate nightly news broadcasts are the opposite of good tidings. Most efforts at prevention and law enforcement seem more like reactive attempts to contain ever expanding problems rather than discernable public triumphs. In recent American history, crime rates seem to increase on the front page and moderate in obscurity.

The recent decline in homicides in New York City is an exception to the usual pattern, the most celebrated example of crime-news-as-good-news in decades. No doubt part of the public attention can be explained because the story took place in the media capital of the United States. But more than location made the New York story newsworthy. The drop in homicides was both large and abrupt--the homicide rate in the nation's largest city fell 52% in five years. Further, changes in police manpower and strategy are widely believed to have contributed to the decline. If this drop can be plausibly tied to enforcement activities, it would be the most conspicuous success of city police deployment policies in the twentieth century.

This article reports our attempt to assess the extent and causes of the five-year decline in life threatening violence in New York City. Part I puts the homicide decline in a variety of statistical contexts, comparing the drop to previous New York experience and to the experiences of other cities in the United States. Part II examines changes in patterns of homicide during the decline in search of clues about causes. Part III examines available data on crime trends and trends in crime-related phenomena over the years when homicide increased and declined. Did many crime categories fall, and by how much? Was the decline concentrated in a few categories or spread evenly across the spectrum of felony crime? What were contemporaneous trends in drag use, demography, and incarceration? Part IV reviews some explanations of the decline.

1. HOW BIG A DROP?

   This section places the five-year decline in New York City homicide in a variety of historical contexts. How large is this drop compared to previous periods in the history of New York City? Is a drop this substantial a typical event in the history of a big city?

   One important element of context for studying the New York City experience is the recent history of homicide rates in the United States. Figure 1.1 provides national level homicide rates by year over the period 1950-95.

   [Figure 1.1 ILLUSTRATION OMITTED]

   After steady to declining trends until the early 1960s, the homicide rate doubled in the period from 1964-74 from just under 5 per 100,000 at the beginning of that period to just under 10 per 100,000 in 1974. After 1974, the trend over time appears to be fluctuations around the 1974 high. Homicide dropped somewhat in the mid-1970s then climbed back up to just above 10 per 100,000 in 1980, dropped in the early 1980s, then climbed back up close to the 1980 high point in the early 1990s only to drop off again after 1991.

   If all the homicide trends since 1964 were forced into a single trend line, the direction of the trend would be upward. If the doubling of rate prior to 1975 is plotted separately, however, the two decades since 1975 represent a trendless fluctuation down from the new peak rate of 10 per 100,000 per year and then back up. By 1995, the homicide rate had dropped back near but not yet below the twenty-year low set in 1984.

   The New York City history can be briefly stated. In its relative and absolute magnitude, the homicide drops after 1992 were by far the largest in the postwar history of New York City. The second largest percentage drop was 25% from 1981-1985. The number of lives involved is even more impressive, with more than 1,100 fewer homicide victims in New York City in 1996 than in 1992. This reduction in homicide far exceeded the total number of homicides the city experienced each year in the 1950s and early 1960s.

   The comparison of this New York experience with the experience of other cities is a more complicated story. Figure 1.2 begins the analysis by reporting on the largest five-year declines in homicides reported since 1945 in each of the fifteen largest cities in the United States as of 1950.

   [Figure 1.2 ILLUSTRATION OMITTED]

   The five-year records for big cities vary from a 61% decline in Pittsburgh through a 15% decline in Chicago. The median decline was 40% and seven cities reported highest decline percentages between 43% and 38%. The New York decline is the third highest for the major cities, behind Pittsburgh, equal to Houston, and about 25% larger than the cluster of city records around 40%. Based on straight arithmetic, the New York experience is not unprecedented, but is a higher percentage drop than twelve of the nation's fifteen biggest cities have experienced in a five-year span.

   There are two other city comparisons that help to illuminate the relative standing of the New York City decline. Figure 1.3 drops the constraint of a five-year span and searches for the largest homicide declines in a decade or less for each of the thirteen largest cities in 1950.

   [Figure 1.3 ILLUSTRATION OMITTED]

   New York City's five-year drop maintains its position as the third largest in this comparison, but a much larger number of cities record declines of the same general magnitude of New York's 51%. It is worth noting, however, that the very largest cities in the United States other than New York City have more modest percentage declines than the smaller cities. New York City's population greatly exceeds Houston's, the next largest city with a similar relative homicide decline.

   Figure 1.4 tests the five cities with the highest five-year declines against the hypothesis that an abnormally high homicide rate for a short period of time is the reason for a large drop in the homicide rate. The low year in the largest five-year decline is compared to the mean rate in that city for the previous fifteen years.

   [Figure 1.4 ILLUSTRATION OMITTED]

   Figure 1.4 helps the observer tell the difference between a big drop that occurs when a city rate returns to near historically normal rates from an atypical high level and declines that still look dramatic when compared with homicide in a city over the long term.

   Three of the five largest declines (including New York City) represent a substantial reduction from the average rate in the previous fifteen years. For Pittsburgh, Houston, and New York City, it is difficult to argue that the drops recorded were merely regression from abnormally high rates. The Boston decline, by contrast, was only to a point 11% below the previous long-term mean. Regression is a much more likely explanation in that case.

   There is one further measure of possible regression relevant to our inquiry. In an era when very wide swings in rates of criminal homicide occur, sharp downward movements in the homicide rate might be expected after sharp upward movements. The theory here is that some downward momentum might be expected after a steep climb in homicide rates for the same reason that rollercoasters tend to drop faster when they have climbed to a great height. The 59% drop in Houston homicides after 1991 came after a sharp increase in rate. Between 1987 and 1991, Houston's homicide rate climbed from 18.5 per 100,000 to 36.3--a 96% increase. Viewed against this background, both the swiftness and the extent of the decline is less astonishing. When measured against the previous low point in 1987, the Houston homicide rate had declined 19.6% instead of 59%, from 18.5 to 14.9 per 100,000. Moreover, the first year in which the Houston homicide was lower than its 1987 level was 1995, three years after the decline started. The real possibility of cyclical variations means that Houston may not have broken out of its long-term pattern until late in the current decline, and that the current homicide level is only modestly lower than the previous low. Similar measures can be found in Dallas, where a 58% decline in homicide brought the 1996 homicide rate to 23% under the city's 1983 rate.

   How does New York City fare on this measure? To use the rollercoaster metaphor, the climb in New York City was not as steep prior to the post-1991 drop, so that less of the New York decline seems likely to be merely cyclical. Between 1985 and 1991, New York City's homicide rate had increased 56% from 17.5 to 27.3 per 100,000. The homicide rate in 1996 was 23% below the 1985 rate, so that about half of the decrease noted by 1996 was outside the range of cyclical variations experienced in the recent past in New York City.

   Still, much of the New York decline could have been cyclical, and this possibility constrains our capacity to explain the homicide decline on other grounds, or to time with precision when New York homicide started behaving in non-cyclical fashion. As late as the end of 1995, the New York homicide rate was within 10% of its 1985 level. The rate of lethal violence broke important new ground only after 1995 or 1996.

   A number of conclusions can be drawn from this brief excursion into the natural history of homicide rates in U.S. cities. The percentage decline experienced in New York in five years is quite large but by no means unprecedented in major cities. The largest five-year drop in the average major city since the Second World War is about 40%--a figure 20% smaller than New York City's. When declines over a six- to nine-year span are allowed, six other cities produce declines of a magnitude similar to the New York experience at some point in post-war history. These declines usually come after periods of general increase in homicide rates. The most common year as a starting point for a city's record decline was 1980, the highest year in U.S. homicide for at least a half century.

   Declines...