1 New York State Department of Correctional Services, Division of Program Planning, Research and Evaluation. Between 1980 and 1995, the number of offenders sentenced to either state or federal institutions grew by about 400 percent. U.S. Department of Justice. Bureau of Justice Statistics. Sourcebook of Criminal Justice Statistics 1996. U.S. Government Printing Office, 1997.
3 Darrel Steffenmeier and Miles D. Harer, “Bulging Prison, an Aging U.S. Population, and the Nation’s Violent Crime Rate,” Federal Probation, Vol. 2, June 1993, pp3-10.
4 Johnnie K. Baxtrom v. R. E. Herold, 383 U.S. 107.
6 Laws of 1973, chapters 276, 277, 278, and 1051.
9 U.S. Department of Justice, op. cit.
10 Quote attributed to Dr. Andrew Karmen, professor of sociology at John Jay College of Criminal Justice, in “Steep Drop in Random Killings Signals Shift in New York,” The New York Times, December 29,1996, p. 25.
11 Comparable data for 1997 are not yet available from the New York State Division of Criminal Justice Services.
12 U.S. Department of Justice. Federal Bureau of Investigation. Uniform Crime Reports. Preliminary Release—January through June 1997. Internet.
13 New York State Division of Criminal Justice Services.
14 “As New York Homicides Fall, Rate of Solved Cases Goes Up,” The New York Times, June 2, 1997, p. B1.
15 See The New York Times, December 20, 1996, p. 1.
16 See Alfred Blumstein, “Youth Violence, Guns, and the Illicit-Drug Industry,” The Journal of Criminal Law & Criminology, Vol. 86, No. 1, 1995.
17 Blanche Frank and John Galea, “Current Drug Use Trends in New York City,” in National Institute on Drug Abuse, Epidemiological Trends in Drug Abuse, Proceedings: Community Epidemiology Work Group, Vol. 2, December 1996, pp 160-173.
18 Reports indicate that there has been no abatement in heroin use. However, many experts acknowledge that heroin users tend to be less violent than crack users. There are also reports that marijuana use is on the rise. However, more than 90 percent of marijuana-related arrests are misdemeanor arrests and, therefore, will not result in sentence to State prison.
19 “Crime Lab; Mystery of New York, the Suddenly Safer City,” The New York Times, July 23, 1995, Section 4.
22 Clearly, individuals who are incarcerated are no longer committing crimes on the street. Just as clearly, the rise and fall in criminal activity clearly has an impact on incarceration rates. Further analysis which disentangles these simultaneously causal effects is necessary before drawing any firm conclusions. For a discussion of this issue, see Thomas B. Marvell and Carlisle E. Moody, Jr., “Prison Population Growth and Crime Reduction,” Journal of Quantitative Criminology, Vol. 10, No. 2, 1994.
24 Frank and Galea (1996), p. 162.
25 A simulation model is a mathematical representation of a real-world system. Simulation models are typically input-output models and, therefore, inherently appropriate for a model of the prison system, where this month’s inmate population is equal to last month’s population, plus the number of new admissions, minus the number released. A computer simulation model provides the researcher with an important tool with which to forecast the future behavior of a complex social system, such as the criminal justice system. Simulation models also permit the analyst to perform simple experiments in order to answer “what if” questions related to policy changes. On the role of simulation in forecasting and policy analysis, as well as criteria for model acceptability, see, for example, Robert E. Shannon, Systems Simulation the Art and Science, Prentice-Hall, 1975.
26 The Committee staff model is “dynamic” in that it demonstrates how the system changes over time on the way to a new steady state due to either a change in one of the inputs or the impact of policy change. It is “disaggregated” in that it divides the prison population into smaller groups (see below). The level of disaggregation is normally based on what level of detail is necessary to produce an acceptable level of accuracy, as well as the nature of the policy changes to be modeled. Finally, the model is considered a “flow” model in that offenders flow from arrest to conviction and sentence to prison, from prison to parole, in some instances from parole back to prison, and so on. Based on these flows and other model parameters, the model calculates the “stock” of offenders that have accumulated in the incarceration phase at a given point in time and reports this number as the inmate population. For a review of the types of models that are typically used for criminal justice system projections, see William Rhodes, “Models of the Criminal Justice System: A Review of Existing Impact Models,” Abt Associates Inc., 1990. “Stock and flow” models have been used to better understand processes as different as water levels in reservoirs, environmental contamination, and school population projections. For a classic textbook treatment of stock and flow models, see Edith Stokey and Richard Zeckhauser, A Primer for Policy Analysis, W.W. Norton & Co., Inc., 1978.
27 On the role of statistical models in computer simulation, see Rhodes (1990). The statistical approach to forecasting offers an objective set of standards by which to evaluate model performance, as well as the accuracy of the forecast. Methods such as ex post simulation and the construction of “goodness of fit” measures, such as the root mean squared error, permit tests of the model’s performance by assessing how closely the model can reproduce the historical data series. How well the turning points in the data are captured is another important standard by which to judge a projection method. On the importance of these criteria for the purposes of forecasting see, for example, Robert S. Pindyck and Daniel L. Rubinfeld, Econometric Models and Economic Forecasts, third edition, McGraw Hill, 1991.
28 More technically, we have estimated a two-equation econometric model. We include autoregressive terms in both equations to allow the model to fully capture all available trend information. To account for the fact that both violent and non-violent arrests are vulnerable to common shocks, we estimate the two equation simultaneously using a method known as seemingly unrelated regression. This method takes into account the cross-equation error covariances, thereby increasing the precision of the estimates. The more precise the estimates, the more confidence one can have in the accuracy of the forecast. A rigorous process of model selection was applied, governed by well-accepted selection criteria. A more detailed discussion of model specification and selection can be found in Pindyck and Rubinfeld (1991).
29 An essential step in the process of simulation model building process is the task of model validation. Properly validating the model reduces the probability that erroneous results will be utilized by policymakers. Two important elements of the validation process are the comparison of simulated data with real system data, as well as the testing of the model’s assumptions, usually involving the use of statistical tests. For example, one possible assumption regarding future arrests might be that the number of arrests per month remain constant at the average level of the past twelve months. This assumption is equivalent to the statement that monthly arrests tend to follow a pattern which statisticians refer to as “white noise,” and that there is no real downward trend. However, statistical testing permits us to rule out the white noise assumption with a very high probability of accuracy. On the importance of model validation and the role of statistical testing, see Shannon (1975).
30 Future parole and conditional release violators and “other” admissions are projected using a two-equation econometric model using the seemingly unrelated regression method.
32 An offender might serve less than six-sevenths of his sentence in State prison if he has been credited with time served in a local jail.
33 During the 1997-98 budget negotiations, the Committee’s population projections for the end of the 2001-02 fiscal year were over 10,000 below the Executive projections for the same point in time. Prior to the presentation of the 1998-99 Executive budget proposal in January, the Executive reduced their projections for both the current year and out-year population. As of the present time, the Committee staff’s population projection for the end of the 2001-02 fiscal year is 2,911 inmates below the Executive forecast.
34 Jonathan P. Caulkins, C. Peter Rydell, William Schwabe and James Chiesa, Mandatory Minimum Drug Sentences. Throwing Away the Key or the Taxpayer’s Money? Rand Corporation, 1997.