Understanding the time path of crime.

• Author: Donohue, John J.

With all of the random factors that influence the amount of criminal conduct, it is virtually impossible to fully explain or precisely predict the crime rate at any point in time. If the World Trade Center bombers had succeeded in their goal of toppling the massive towers, tens of thousands could have died in New York that one day in 1993. The nation's annual murder rate would have doubled or tripled from one incident. Fortunately, such extreme catastrophes are rare--the comparatively minor but still horrific killing of 168 residents of Oklahoma City in 1995 was too small against the background of 21,600 murders across the nation to have a pronounced effect on the U.S. murder rate in that year.(1) There will always be random events that escalate the rate of crime, such as innovations in illegal drug markets, and, similarly, random events that tend to depress it, such as bad weather (it keeps the criminals at home), charismatic religious leaders, or widespread reductions in enthusiasm for illegal drugs.

The primary goal then, in understanding the long-run trends in crime and what is likely to happen in the future, is to take one's focus off the short-term fluctuations so one can identify the stable long-run patterns. Distinguishing stable trends from temporary fluctuations is essential to understanding how crime is affected by changes in criminal justice policy, as well as by varying social, economic, and demographic influences. If we confuse fluctuations with trends, our predictions of where crime is headed can be wildly inaccurate, and our search for causal explanations of recent patterns can be very misleading. For example, as we reflect upon the exciting and salutary recent sharp drops in crime, it makes a huge difference whether these are just temporary improvements varying around a long-term unchanged trend or the signal of a precipitous and sustained improvement from the previous pattern of slow decline that has been operating for two decades. Similarly, in trying to explain the very recent sharp drops in crime, one must exercise care in attributing causal significance to forces, such as the sharply rising rate of incarceration, that have been operating relentlessly for a quarter of a century.

This paper, then, will attempt to sort out the long-term trends in crime over the last fifty years from the short-term fluctuations around those trends. As we will see, there have been two clear long-run trends in crime over the last half century: one involving sharply rising crime until the late 1970s, followed by the second, a period of slow decline over the next two decades. As one might expect, there have been considerable short-term fluctuations around the two long-run trends, and indeed, the later period has experienced greater variability in crime around the long-term declining trend than had been the case during the initial period of the rising secular, or long-term, trend in crime.(2) Part I documents these broad patterns, and discusses how they illuminate the issues of why crime has fallen and where it is likely to be headed in the future. Part II builds upon this discussion to show that increased levels of incarceration and favorable demographic shifts contributed to the slow decline in crime over the last two decades, but cannot explain the sudden drop in crime in the mid-1990s after the abrupt increases in crime of the late 1980s. The reasons for the short-term fluctuations are probed and various positions that were advocated during the conference are evaluated. Part III concludes by noting that the growing cost of incarceration suggests that, at some point, the public will call for an end to further increases in the number of prison inmates. Since increasing incarceration, more police, and favorable demographics have been modestly offsetting the influences pushing towards higher crime, when the increases stop and the demographic trends turn unfriendly (as they now have), crime will begin a slow secular rise for the first time in two decades, unless some other force (better policing strategies, effective social programs) controls crime or the unknown long-term criminogenic forces in society (the breakdown in the family, pernicious media influences, declining schools, growing drug use and drug markets) abate.

1. THE PATTERN OF HOMICIDES OVER THE LAST HALF CENTURY

   1. DISTINGUISHING LONG-TERM TRENDS AND SHORT-TERM FLUCTUATIONS

      Because of the poor quality of the data published by the FBI over the last half century, it is very hard to provide a comprehensive and accurate assessment of the long-run patterns of all aspects of crime. It is possible, though, to focus on the one crime--murder--that is well measured, and for which a reliable long-term time series can be created. While homicide data may not be perfectly reflective of the time trend in all crimes, it does seem to follow the pattern of most other street crimes fairly well during the recent periods when more accurate data is available for these other crimes.(3) Thus, while murder may not be a perfect proxy for crime, it is simply the best we have. For the rest of this paper, then, I will rely on murder data to define and explain the broad patterns in crime over the last five decades. Figure 1 plots the national homicide rates from 1950 through 1997.

      [Figure 1 ILLUSTRATION OMITTED]

      National crime patterns can be thought of as being composed of long-term trends and short-term fluctuations around these long-term trends. Figure 1 also plots two trend lines (the predicted murder rates), which reveal that from 1950 to 1977 the murder rate rose at an annual rate of 4.4%, and from 1977 through 1997 it fell at a rate of roughly 0.6%.(4) At the same time, there has been substantial deviation around the trend. Over this forty-eight year period, the two predicted homicide trend lines explain almost two-thirds of the variation around the mean national murder rate of 7.32 per 100,000. Thus, analysts need to find explanations for both the two long-term trends-one strongly adverse for the period before 1977, and one mildly benign for the subsequent period--as well as for the variations above and below these trends.

      This descriptive scheme aids in the process of explanation since it clarifies the need to find long-term explanations for the steady long-term trends in crime, and more episodic and variable explanations for the short-term variations about the long-term trends. For example, if one is trying to explain the post-1977 downward trend in crime, the sustained increase in incarceration rates over this period is certainly part of the explanation. At the same time, this steady increase in incarceration is not a reasonable explanation for the sharp drop in crime that has occurred since 1993.(5)

   2. WE CAN'T KNOW WHERE CRIME IS HEADING WITHOUT KNOWING WHY IT HAS FALLEN

      Perhaps the most important lesson from Figure 1, though, is that the excitement over the prospect of a sustained sharply lower crime rate may be premature. Posit, for a moment, that the post-1977 linear predicted homicide rate truly represents the current long-run trend in crime. In this event, we can obviously be happy that crime is headed downward, but the euphoria of the last few years must be tempered by the realization that the slope of the long-term downward trend is obviously much more gradual than that of the recent drop. We would achieve an enormous public policy victory if we could engineer a return to the low crime rates of the 1950s and early 1960s But if the post-1977 linear time trend shown in Figure 1 accurately reflects the long-run trend in homicide, then it will be a very long time before that goal will be reached.(6)

      I do not mean to be pessimistic, and I would love to think that we have entered a third phase in the post-World War II pattern of homicide, in which homicide rates will now be falling more sharply into the future. There are certainly reasons why any trend in crime tends to gain momentum.(7) But as I look at Figure 1, I cannot rule out the possibility that the last five years are similar to the first five years of the 1980s. If so, then we might expect that the present period of rapidly declining murder rates will be followed by an increase, as the national homicide rate returns to its more gradual long-run pattern of decline. In a few years, of course, we will have a better answer to this question: if the crime rate continues to fall at its current rate, then it will likely mean that there has been some shift in the fundamentals of homicide, instead of a benign short-term variation about the unchanged post-1977 long-run trend.(8)

      1. Alternative Time Paths of Crime

         Of course, there is nothing set in stone about the way that I have modeled the last half century of homicide rates as having an upward trend before 1977 and a downward trend thereafter. I simply allowed the data to have one break in the series, and then found the line of best fit across all possible years in which the break could occur. Thus, the statistical data indicate that if there is to be only a single break in the time trend, this break occurred in 1977. But two qualifications should be mentioned. First, it is conceivable--as some commented at the conference--that homicide rates over the last fifty years are better explained by a single curvilinear trend rather than by the two linear trends shown in Figure 1. Figure 2 graphs just such a quadratic equation.

         [Figure 2 ILLUSTRATION OMITTED]

         If Figure 2 depicts the true core pattern of homecides, then it is apparent that homicide rates will be restored to the low levels of the early 1960s fairly quickly. In fact, Figure 2 predicted the homicide rate would fall to the 1965 level within five years.(9) The reason for this sharply different prediction is that the predicted homicide-rate curve in Figure 2 gives far greater weight to the observed sharply downward trend in crime over the last few years. But visual inspection suggests that the predicted...