DECLINING VIOLENT CRIME RATES IN THE 1990S: Predicting Crime Booms and Busts.

INTRODUCTION

Recent declines in rates of violent crime in the United States caught many policymakers and researchers off guard. These declines were perhaps more surprising in that they came on the heels of dire predictions about the rise of a generation of "superpredators" who would soon unleash the full force of their destructive capacities on an already crime-weary nation (Bennett et al 1996, Fox & Pierce 1994).

I argue in this paper that changes such as those involving the recent downturn in violent crime rates usually surprise researchers because we rely mostly on cross-sectional analysis and rarely study social phenomenon like crime trends in a longitudinal context. I begin this article by looking more closely at the recent downturn in violent crime rates, comparing it to trends since World War II. I then summarize and evaluate several common explanations for changing violent crime rates. Finally, I identify a set of concepts that might be useful for moving our thinking about social phenomenon such as crime toward a more longitudinal framework.

Studying Crime Trends in Postwar America

I concentrate here on violent crime trends for the United States. While violent crime generally includes murder, rape, robbery, and aggravated assault, researchers agree that data are probably most accurate for murder and robbery (Gove et al 1985, O'Brien 1996). A focus on the United States is strategic in that it accounts for the greatest volume of research, although such a focus simultaneously underscores the strategic importance of developing more cross-national comparisons.

Researchers interested in studying violent crime trends in the United States are limited to two main data sources: the Uniform Crime Reports (UCR), collected annually since 1930 by the Federal Bureau of Investigation; and the National Crime Victimization Survey (NCVS), collected every six months since 1973 by the Bureau of Justice Statistics (Rand et al 1997). Space limitations prevent a detailed assessment of the strengths and weaknesses of these data sets (Biderman & Lynch 1991, LaFree 1998a:Ch. 2). I concentrate here on the UCR because it provides the longest uninterrupted crime series data for the United States.

Figure 1 shows UCR trends for murder and robbery, 1946 to 1997. I begin the series in 1946 because of serious validity problems with pre-World War II UCR data. I use "longitudinal" throughout this paper to refer to studies based on annual time-series analysis.