Understanding Crime Trends: Workshop Report (2008)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

1 Introduction Richard Rosenfeld and Arthur S. Goldberger Changes over time in the levels and patterns of crime have significant consequences that affect not only the criminal justice system but also other critical policy sectors. Yet compared with such areas as health status, hous- ing, and employment, the nation lacks timely information and comprehen- sive research on crime trends. Consider a recent example. After declining or remaining stable for over a decade, violent crime rates rose in many American cities in 2005 and 2006. What is known about these changes? What brought them about? Could they be anticipated? The honest answers are: very little, no one knows, and no. Descriptive information and explanatory research on crime trends across the nation that are not only accurate but also timely are pressing needs in the nation’s crime-control efforts. Without useful and reliable information, national and local policy makers fly blind when formulat- ing and evaluating the effectiveness of policy interventions or, as a recent National Research Council (NRC) report on criminal justice evaluation research observes, they must generate ad hoc outcome indicators for each new policy assessment (National Research Council, 2005, pp. 59-60). In the absence of common and timely crime indicators, the U.S. Department of Justice sent teams of “auditors” to a select group of cities in 2007 to examine local records and consult with law enforcement officials concern- ing local crime trends. Although there may be good reasons for proceeding in this way, the contrast with using available employment indicators to evaluate local labor market conditions is illuminating. Faced with concerns about a rise in unemployment, the U.S. Department of Labor would not have to, at least in the first instance, send auditors across the country to  

 UNDERSTANDING CRIME TRENDS examine local employment records. That information is routinely compiled, stored, and updated in centralized records systems that can be accessed by local, state, and national officials, researchers, the press, and the public. These information systems constitute the basis of a policy evaluation infra- structure that is indispensable for assessing economic policy in the United States. Comparable evaluation capabilities do not exist in the policy area of criminal justice (Rosenfeld, 2006). Monitoring Crime Trends The nation lacks a comprehensive, coherent, and up-to-date infra- structure to monitor crime trends and relay the resulting information to law enforcement agencies, researchers, policy makers, and the public. The federal government sponsors two major crime data collection programs, the Uniform Crime Reports (UCR) of the Federal Bureau of Investiga- tion (FBI) and the National Crime Victimization Survey of the Bureau of Justice Statistics. Although both programs provide essential information about crime levels, patterns, and trends, neither does so with the speed and level of detail necessary to inform local law enforcement planning or state and national policy responses to emerging crime problems. The pau- city of high-­quality information for criminal justice planning and research stands in sharp contrast to the rich assortment of monthly and quarterly data collections on specific topics that characterizes other policy areas. The rapid and continuous public dissemination of economic data and forecasts (“economists predict slowdown in next quarter,” “monthly housing starts off by 2 percent”) has no equivalent in criminal justice, even though timely information on changes in serious violent and property crime rates is just as vital to the nation’s health and welfare as information on changing levels of industrial production or consumer prices. The FBI’s UCR do provide national and local crime indicators, but the UCR data are released several months after the relevant reporting period. The year-end 2005 data, for example, were not released until September 2006, and the preliminary data for the first six months of 2006 were not available until December of that year (http://www.fbi.gov/ucr/ucr.htm). Even though the time lags have been shortened in recent years, they often extend well beyond the planning and response horizons of local jurisdic- tions. Without timely crime indicators to draw on, Justice Department officials were caught short when police chiefs and mayors from around the country assembled in Washington, DC, in August 2006, to demand federal assistance in combating local crime increases (Files, 2006). The Police Exec- utive Research Forum (PERF) later reported sharp crime increases on the basis of a survey of the cities represented at the August meeting (Johnson, 2006; the PERF report is available at http://www.­policeforum.org). But the 

INTRODUCTION  representativeness of the survey results is uncertain, and they offer only a snapshot of local crime patterns rather than the kind of continuous moni- toring needed for rapid and ongoing policy assessment and intervention. THE NRC WORKSHOP In April 2007 the NRC held a two-day workshop to address in a preliminary way key substantive and methodological issues underlying the study of crime trends and to lay the groundwork for a proposed multiyear NRC panel study of these issues. Six papers were commissioned from leading researchers and discussed at the workshop by experts in sociology, criminology, law, economics, and statistics. The authors revised their papers based on the discussants’ comments. In accordance with standard NRC procedures, the papers were sent out for external review and revised again. The six final workshop papers are the basis of the current volume. The workshop committee was necessarily selective regarding the range of topics that could be covered in a two-day workshop. The committee asked Alfred Blumstein and Richard Rosenfeld to summarize changes in rates of serious crime in the United States over the past several decades and identify factors that may be driving those changes. Karen Heimer and Janet Lauritsen were asked to address trends in victimization and offending by sex. Jeffrey Fagan addresses the prospects and challenges of analyzing neighborhood-level crime trends. Two researchers, Eric Baumer and John Pepper, were asked to perform separate analyses, including forecasting crime rates, on a city-level dataset specially created for the committee by Robert Fornango of Arizona State University. Finally, the committee asked Steven Durlauf to discuss statistical and theoretical issues in drawing causal inferences from observational data on crime rates. The workshop was intended to highlight outstanding issues in the a ­ nalysis of crime trends rather than to develop a consensus agenda for research, let alone offer consensus recommendations for policy makers. Thus the chapters do not form an integrated whole but rather an explo- ration of the field. Although the six papers cover a diverse assortment of substantive and analytical issues, clearly they do not exhaust the range of relevant topics of interest to the scientific community or policy makers. For example, no effort was made to draw on trends in other countries. None of the papers deals specifically with the problem of race, ethnicity, and crime. That issue, as well as the related question of the impact of immi- gration on crime trends, requires much more extensive evaluation than is possible in a workshop format and could be a major topic to be considered by an NRC panel study of crime trends (see Peterson, Krivo, and Hagan, 2006, for a recent treatment). Nor did the workshop address the research and policy implications of trends in so-called white-collar crime or in the 

 UNDERSTANDING CRIME TRENDS multiple forms of cyber crime. Little systematic research exists on the use of the Internet as a conduit for stolen goods, child pornography, identity theft, or fraud. Building the science in this important area of public concern might also be on the agenda of any future panel study of crime trends to emerge from this workshop. RESEARCH CONTEXT Rates of serious violent crime in the United States have exhibited marked fluctuations over the past 30 years (see Blumstein and Rosenfeld, this vol- ume). Homicide and robbery rates rose to peaks in 1980 and the early 1990s and fell by over 40 percent through the end of the century, during what one analyst has termed “the great American crime decline” (Zimring, 2006). Social scientists, policy makers, and law enforcement officials were caught off guard by these changes, in part because they did not become apparent in the available statistical indicators until well after they had begun. Little planning for spikes in resource demands or timely responses are possible under such circumstances. The local police department is perhaps the last remaining complex organization in American society without the capacity to anticipate and plan for changes in its environment, yet it is assigned virtually all of the responsibility for responding to crime increases and more than its share of the blame when the responses prove inadequate. Although research on crime trends remains quite limited, the emerg- ing assessments of the 1990s crime decline yield some useful insights on which more comprehensive and sustained efforts can build (Blumstein and W ­ allman, 2005; Rosenfeld, 2004; Zimring, 2006). The first lesson from that research is that disaggregating total crime rates by age, race, sex, city size, and other factors reveals important differences in the timing, mag- nitude, and duration of group-specific trends. The second lesson is that multiple causes underlie the crime drop and, by extension, longer term variations in crime rates. Some progress has been made in identifying can- didate explanatory factors, which include the quadrupling of the nation’s prison population since 1980; cyclical variations in unemployment, wages, and other economic conditions; and the changing dynamics of illegal drug markets (Blumstein and Wallman, 2005). But the relative contribution of these factors to the 1990s crime drop or more recent local-level crime trends remains uncertain (Levitt, 2004; Rosenfeld, Fornango, and Rengifo, 2007). Nor is it clear whether they are themselves sufficiently predictable to serve as the basis for useful crime forecasts. Another lesson from recent research is the value of distinguishing short- from long-run changes in crime rates and in the factors hypothesized to explain those changes. Consider how changes in the economy may affect crime over the long and short run. A well-known account suggests that the 

INTRODUCTION  move from a manufacturing-based to a service-based economy during the latter half of the 20th century contributed to an increase in crime among population groups, young minority men in particular, lacking the train- ing and skills needed for the better paying jobs in the rising service sector (Wilson, 1987, 1996). Cities hit hardest by deindustrialization experienced large and sustained crime increases (Parker, 2004). But deindustrialization and the social changes attributed to it, such as the growth of “oppositional cultures” among minority youth in distressed urban areas (Anderson, 1999) are, by themselves, unlikely to account for short-run swings in crime rates, such as those occurring in some cities in 2005 and 2006. Better candidate explanations include cyclical economic changes, prison admissions and releases, local enforcement initiatives, and other factors subject to year-to-year fluctuation, some of which may acti- vate local grievances or more widespread and long-standing psychological or cultural conditions. A final and related insight from the recent research on crime trends distinguishes sources of change in crime rates that apply across local areas from those that may be specific to particular jurisdictions. Some factors, such as economy-wide changes in unemployment or wage rates, poten- tially affect crime conditions across the nation, whereas others, such as gang conflicts or local changes in policing strategies, influence crime rates in some places but not others. A good example of the importance of this distinction is the dramatic fall in New York City’s crime rates during the 1990s, at roughly double the rate of decline registered at the national level (Zimring, 2006). The difference suggests that New York’s crime drop is related to distinctive local conditions as well as factors common to other cities. One candidate for explaining New York’s crime reduction “bonus” is its CompStat initiative and accompanying order-maintenance enforcement strategies begun in the early 1990s. New York clearly differs from other cities in other respects that may have influenced its crime trends, and recent research has found either small or no effects of CompStat on the New York City crime drop (Fagan, this volume; Harcourt and Ludwig, 2006; R ­ osenfeld, Fornango, and Rengifo, 2007). THE CURRENT VOLUME Four lessons from the emerging research on the crime drop— (1) disaggregating crime rates, (2) identifying multiple sources of variation, (3) distinguishing long swings from year-to-year variation, and (4) differen- tiating general and specific changes—form the context for the contributions to this volume. None of the chapters addresses all four of the issues but taken as a whole the volume illustrates the importance of each of them. In Chapter 2, Alfred Blumstein and Richard Rosenfeld underscore the value 

 UNDERSTANDING CRIME TRENDS of disaggregating recent crime trends by age, race, and ethnicity to disclose the multiple and differing factors associated with group-specific variation in trends. Their chapter identifies factors that had been established by prior research to be associated with changes in crime rates, such as demographic shifts, growth in incarceration, drug markets, and changing economic con- ditions. They identify other factors, such as policing innovations, firearm availability, street gangs, childhood socialization, and investments in social services that may influence crime trends but for which the existing evidence is fragmentary or inconsistent. The authors conclude that developing sound empirical explanations of past crime trends is an important means of improving the capacity to make informative and reasonably accurate forecasts of future changes. They sug- gest that only with a large investment of resources can criminologists hope to make their forecasting models as worthwhile as those produced in other disciplines. To be most useful, it is also important to differentiate efforts at generating national, regional, and local (for a particular city or neighbor- hood) crime trend estimates. Blumstein and Rosenfeld conclude that the influence of research on policy will be limited in the absence of a substantial upgrading in the nation’s capacity to monitor crime trends. That will require additional resources devoted to compiling, disseminating, and updating the data. They suggest that the National Institute of Justice can play an important part in this process by establishing an ongoing research program devoted to analyzing crime trends. In Chapter 3, Karen Heimer and Janet Lauritsen use data from the National Crime Victimization Survey to examine changes between 1980 and 2004 in female and male violent offending and victimization and victim-offender relationships in violent incidents.  This work has not been done previously except for homicide, and so the chapter constitutes a unique contribution to the field.  Among the authors’ findings are a widen- ing of the gender gap in intimate partner homicide victimization due to a greater decline in victimization among men, a narrowing of the gender gap in overall violent offending, an increase in the proportion of assaults involv- ing female victims, and an increasing likelihood of female involvement in violent interactions, both as perpetrators and as victims. The authors note that the modal category of violent crime in 2004 is not the same as it was in earlier decades. Recent declines in stranger violence have been such that, by 2004, male victimization by strangers was no longer greater than male or female victimization by nonstrangers. Why stranger violence has declined more rapidly than nonlethal acquain- tance violence is a challenging question for future research.  In addition, although female violent offending against strangers has always been low, the large percentage increase in such violence that occurred before violence 

INTRODUCTION  rates peaked in the mid-1990s warrants further investigation.  The authors conclude that nonstranger violence remains a critical part of violence in the United States, and women and men are now affected equally by violence by acquaintances. At the same time, they caution that violence prevention strategies may not work equally well for men and women and that gender differences in the effectiveness of interventions to reduce violence should be evaluated systematically. In Chapter 4, Jeffrey Fagan reviews research on factors that influence changes in crime rates between and within neighborhoods in cities over time. He examines local area studies of neighborhood and crime, focusing on neighborhood structures and processes. The study of neighborhoods has stimulated a rich body of sociological theory to conceptualize space and its effects on individuals and populations. But studies of the impact of neigh- borhood change on crime have been rare and are usually limited to a few neighborhoods in single cities. Fagan identifies several challenges to theory, measurement, and analysis that affect estimates of why and how neighbor- hood crime rates change. He offers a causal account that frames changes in violence within and between neighborhoods as contagion and diffusion processes. The challenges are illustrated with data from a panel study of violent crime in New York City neighborhoods from 1985 to 2000. Fagan notes the difficulties associated with compiling systematic data on these issues and calls for building an infrastructure in cities for neighbor- hood data to support research on neighborhoods and crime. To overcome some of the political challenges to achieving this goal, he calls for the shift- ing of social and professional norms toward more open and transparent data systems to monitor changes in local crime rates that mirror changes in each city’s neighborhoods. The chapters by Eric Baumer and John Pepper analyze crime trends in U.S. cities between 1980 and 2004. The units of analysis are 240 large U.S. cities with populations of 100,000 or more according to the 2000 census. The data consist of rates of homicide, robbery, burglary, and motor vehicle theft as measured by the FBI’s Uniform Crime Reports. The data also include annual measures of drug arrests, state-level incarceration rates, the number of police per 100,000 population, and demographic, social, and economic indicators drawn from the 1980, 1990, and 2000 censuses. Each researcher was free to augment the common dataset with additional indicators. The committee asked them to use their analysis of 1980-2004 crime trends to forecast crime rates into 2005. In Chapter 5, Baumer develops a comprehensive assessment that sig- nificantly expands the typical set of factors considered in crime trends research. He includes in his models most of the major factors identified in prior cross-sectional and longitudinal research on crime rates. Advocates for the role of particular factors, such as policing, incarceration, abortion, 

 UNDERSTANDING CRIME TRENDS or immigration, have drawn relatively strong empirical conclusions, which ignore other factors that may be equally or more relevant. Baumer high- lights five noteworthy findings from his analysis: 1. Changes in incarceration and crime are significantly related dur- ing the period under consideration. Increases in state prison com- mittals per 100,000 residents tend to reduce crime the following year, whereas increases in the number of persons released from state prisons per 100,000 residents tend to increase crime the next year. 2. Policing variables yield inconsistent findings. A measure of pub- lic order and weapons offenses is unrelated to crime rates, but increases in police force size and the certainty of arrest are associ- ated with crime reductions. 3. Overall results point to a relatively limited role for short-run changes in the economy. 4. Evidence for an association between guns or drugs and recent crime trends is mixed. Alcohol consumption trends do not appear to influence recent crime trends. Firearm prevalence is significantly associated with homicide but not other crimes. Indicators of change in crack cocaine use and market activity exert significant effects on recent crime trends, albeit in somewhat inconsistent ways across measures and crime types. 5. Some of the demographic variables—age, cohabitation rates (but not marriage), percentage of the youth cohort born to teenage mothers—exert significant effects on crime trends. Regression results—coefficients, standard errors, etc.—are not them- selves indicators of the influences of factors on past trends. Changes in the values of the corresponding explanatory variables need to be considered as well. Doing so to estimate the relative contributions of the explanatory variables to observed change in crime rates, Baumer’s analysis supports the conclusion that the rise in youth firearm violence, robbery, and some forms of auto theft during the 1980s can be attributed to the emergence and proliferation of crack cocaine markets. In addition, consistent with other reports, the analysis indicates that lethal violence would have increased even more during the 1980s had it not been for a substantial increase in levels of incarceration and a considerable decline in the relative size of the youth population ages 15-24. Incarceration also emerged as a primary contributor to the decline in burglary and adult homicide during the 1990s, accounting for more than half of the observed declines in both of these crimes. While Baumer’s models fit the sample period reasonably well, his fore- casts diverge substantially from the observed 2005 crime rates. He con- 

INTRODUCTION  cludes that empirical literature on crime trends is in the early stages of development, and much more research is needed before confident conclu- sions can be provided. In Chapter 6, Pepper takes a different approach using several time- series regression models to forecast crime rates. While Baumer included once-lagged crime rates along with a long list of covariates, Pepper focused on the lagged rates, supplemented in part by a very short list of covariates. He examines the possibility of predicting a crime rate series from its past history, thus treating forecasting as distinct from causal analysis. He first illustrates his approach using national data and then turns to the city-level database. He compares the performance of basic panel data models with and without covariates and with and without lags. He also compares two naïve models, one in which the forecast equals the city-level mean or fixed effect, and one in which the forecast equals the last observed rate (random walk forecast). He examines the basic plausibility of the models as well as their prediction accuracy and bias over 1-, 2-, 4-, and 10-year forecast horizons. Pepper found that the forecast models are fragile, in that seemingly minor changes to a model can produce qualitatively different forecasts. Naïve models do relatively well for short-term forecasts, but forecasts are invariably error ridden around turning points, especially, he speculates, when these movements are largely the result of external events that are themselves unpredictable. In Chapter 7, Steven Durlauf, Salvador Navarro, and David Rivers examine the use of aggregate regressions—of the sort run by Baumer and Pepper—as a basis for informing policy decisions. Starting with a formula- tion at the level of an individual, the authors indicate how the individual- level model can be aggregated to produce crime regressions of the type found in the literature. They demonstrate some of the limitations of these regressions, focusing particularly on how empirical findings may be over- interpreted when the link between aggregate data and individual behavior is treated casually. They then discuss the analysis of policy counterfactuals, consider issues of model uncertainty in crime regressions, and illustrate these arguments in the context of two prominent papers in the capital pun- ishment and deterrence literatures. The authors note that assumptions are embedded in any scientific approach, and they attempt to clarify various assumptions needed to inter- pret aggregate crime regressions in terms of individual behavior. They out- line ways of using model-averaging methods and statistical decision theory to broaden the basis of forecasts. (Baumer also discusses the problem of model uncertainty.) The NRC workshop committee asked several researchers to comment on the papers prepared for the workshop: Philip Cook, John Donohue, 

10 UNDERSTANDING CRIME TRENDS Rosemary Gartner, Lauren Krivo, Kenneth Land, Daniel Nagin, Robert Sampson, Justin Wolfers, and Franklin Zimring. Although space limitations precludes publication of the comments in this volume, each of the chapters has benefited from the discussants’ criticisms and suggestions, some of which stand as original contributions to the emerging body of research on crime trends. To take one example, Wolfers suggests that “prediction mar- kets” of the kind used to forecast economic changes and election outcomes may have a useful role in forecasting crime rates (Wolfers and Zitzewitz, 2004). Given the poor forecasting track record of criminologists (Land and McCall, 2001, 2006), some experimentation with differing forecasting methods seems warranted. From this summary of the chapters, it is clear that the current volume does not offer an integrated whole, but rather explores the range of issues that will be addressed as progress is made. What is clear is that progress will require a substantial improvement in infrastructure in the form of cur- rent and comprehensive databases. Past empirical and theoretical analysis has focused on cross-sectional data, sometimes over a few periods. What emerges from this work are correlates of crime rates, which may not be the ones that are relevant for temporal analysis. For example, ethnic composi- tion and income distributions may have strong associations with crime rates across localities at a point in time, yet vary so little over time that they cannot have contributed to the time paths seen. The NRC workshop and this resulting volume represent some of the most serious thinking and research on crime trends currently available. But they also reveal how far there is to go in improving information and research in order to provide useful policy guidance. The hope is that the current volume will stimulate other social scientists to contribute fresh insights and innovative methods to the study of crime trends. REFERENCES Anderson, Elijah. (1999). Code of the street: Decency, violence, and the moral life of the inner city. New York: Norton. Blumstein, Alfred, and Joel Wallman (Eds.). (2005). The crime drop in America (2nd ed.). New York: Cambridge University Press. Files, John. (2006). Police chiefs want U.S. aid in crime fight. New York Times, August 31. Available: http://www.nytimes.com/2006/08/31/us/31crime.html?_r=1&adxnnl=1&oref= slogin&adxnnlx=1219237826-uj+VjluE1+WysknxZi+/hg [accessed August 2008]. Harcourt, Bernard E., and Jens Ludwig. (2006). Broken windows: New evidence from New York City and a five-city social experiment. University of Chicago Law Review, 73, 271-320.   Wolfersand Zitzewitz also submitted an unpublished paper subsequent to a discussion at the 2007 workshop. 

INTRODUCTION 11 Johnson, Kevin. (2006). Dozens of cities see jump in murders, robberies. USA Today, October 12. Available: http://www.usatoday.com/news/nation/2006-10-12-violent-crime_x.htm [accessed August 2008]. Land, Kenneth C., and Patricia L. McCall. (2001). The indeterminacy of forecasts of crime rates and juvenile offenses. In National Research Council, Juvenile crime, juvenile jus- tice. Panel on Juvenile Crime: Prevention, Treatment, and Control, Joan McCord, Cathy Spatz Widom, and Nancy A. Crowell (Eds.). Committee on Law and Justice, Division of ­Behavioral and Social Sciences and Education. Washington, DC: National Academy Press. Land, Kenneth C., and Patricia L. McCall. (2006). Uncertainty in crime forecasts: A recon- sideration. Paper presented at the meeting of the American Sociological Association, Montreal (August). Levitt, Steven D. (2004). Understanding why crime fell in the 1990s: Four factors that explain the decline and six that do not. Journal of Economic Perspectives, 18, 163-190. National Research Council. (2005). Improving evaluation of anticrime programs. Committee on Improving Evaluation of Anti-Crime Programs, Committee on Law and Justice, Divi- sion of Behavioral and Social Sciences and Education. Washington, DC: The National Academies Press. Parker, Karen F. (2004). Polarized labor markets, industrial restructuring and urban violence: A dynamic model of the economic transformation and urban violence. Criminology, 42, 619-645. Peterson, Ruth D., Lauren J. Krivo, and John Hagan. (2006). The many colors of crime: Inequalities of race, ethnicity, and crime in America. New York: New York University Press. Rosenfeld, Richard. (2004). The case of the unsolved crime decline. Scientific American, February, 82-89. Rosenfeld, Richard. (2006). Connecting the dots: Crime rates and criminal justice evaluation research. Journal of Experimental Criminology, 2, 309-319. Rosenfeld, Richard, Robert Fornango, and Andres Rengifo. (2007). The impact of order- m ­ aintenance policing on New York City robbery and homicide rates: 1988-2001. Crimi- nology, 45, 355-383. Wilson, William Julius. (1987). The truly disadvantaged: The inner city, the underclass and public policy. Chicago: University of Chicago Press. Wilson, William Julius. (1996). When work disappears: The world of the new urban poor. New York: Knopf. Wolfers, Justin, and Eric Zitzewitz. (2004). Prediction markets. Journal of Economic Perspec- tives, 18, 107-126. Zimring, Franklin E. (2007). The great American crime decline. New York: Oxford University Press.