SUPPORTING STATEMENT Part A_081512

SUPPORTING STATEMENT

The National Corrections Reporting Program (NCRP)

The Bureau of Justice Statistics (BJS) is requesting clearance to conduct the National Corrections Reporting Program (NCRP) through October 31, 2015. Through the NCRP, BJS collects administrative records on annual movements of offenders in four cohorts: Admitted into prison, released from prison, held in prison at yearend, and discharged from parole (or post-custody supervision). Since 1983 through the NCRP, BJS has annually reported on the movements of offenders through state prison systems. These statistics are part of BJS’ core corrections statistics, as they contribute fundamentally to BJS’ mission of describing transitions and movements of offenders through the criminal justice system.

Specifically, BJS uses the NCRP data to describe changes in the composition and factors affecting the growth of state prison and parole populations. These are issues that have been at the forefront of discussions of corrections policy for decades. For example, during the 1990s, BJS used the NCRP data to help demonstrate the contribution of time served to the increase in the size of the prison population. Since 2000, as prison population growth stabilized and has started to decline, BJS has used the NCRP to help in demonstrating the contribution of decline in commitments from the courts to declines in prison population growth. By virtue of the combination of demographic and offense data collected in the NCRP, BJS has been able to document changes in racial disparities associated with drug offenders in prison. These are but a few of the uses.

In the NCRP state departments of corrections provide individual-level data for each prisoner in their system that meets the definitions provided by BJS. The data in each cohort contains a common core of variables, and each cohort other than the admissions cohort includes additional variables pertaining to the stage in the corrections system process represented by the cohort. The core variables collected for all four cohorts include measures of:

• The date of admission;

• The type of commitment (e.g., from the court, that is, a new court commitment stemming from a felony conviction versus entry as a parole or conditional release violator);

• The offenses for which offenders have been sentenced to prison;

• The lengths and types of sentence imposed;

• Time served in jail prior to admission;

• Demographic attributes of offenders, such as age, race, gender, and education level completed; and

• The entity responsible for offenders (such as the state maintaining jurisdiction over an offender).

In addition, the prison stock cohort includes the common core plus measures of expected (or projected) dates of release from prison. The prison release cohort also includes actual dates of release, actual time served, and method of release from prison (e.g., conditional release onto parole or unconditional release). The parole discharge dataset—which pertains only to offenders released from prison conditionally—also includes data on the type of discharge from parole (e.g., success or failure) and dates of discharge.

BJS has worked to increase state coverage of NCRP data. In 2010, 40 states submitted prison admission records to NCRP, 39 submitted prison release records, and 34 provided parole discharge records. Participation in the collection of year-end prison population records reached its highest level in 2010 since this effort began in 1999, with 36 states contributing data. In addition, NCRP data from earlier years were collected from 8 states. These submissions represent 90.2%, 92.6%, and 81.7% of the total state prison admissions, releases, and year-end stock population in 2010, respectively.

During the past three years, BJS has implemented improvements to the NCRP and during the upcoming three years, BJS plans to implement enhancements. The improvements implemented during the past three years have focused on assessing the reliability of NCRP records, linking records within NCRP to better track movements from admission through release from prison, enhancing the scope of substantive issues that can be addressed with the NCRP, and addressing methodological and estimation issues related to characterizing prison population movements. These improvements are discussed in Section A2, Needs and Uses, below.

The planned enhancements during the upcoming years will focus on improving tracking of movements of offenders between states and better measurement of the terms or periods on post-custody supervision (commonly called parole). These efforts will improve BJS’ capacity to report on recidivism as measured by return to prison, to measure movements of prisoners among states, and to routinely capture information about the veteran status of prisoners.

A. Justification

1. Necessity of Information Collection

The scale, purposes, and impacts of corrections in the United States and in particular of prison populations are ongoing national policy issues. Imprisonment, the nation’s most serious punishment for crime short of execution, is its most costly sanction. From the early 1980s through 2010, the number of inmates held in the custody of states’ departments of corrections increased by three and one-half times, from 371,000 to 1.3 million and over the same period the costs to incarcerate state prisoners increased nearly four-fold (in real dollars) to nearly $40 billion in 2010. The sizable investments by states in recent years in managing prison populations have led to renewed calls about the purposes and benefits of incarceration, and these debates have only intensified as states’ budgets have suffered in the current economic climate.

The central questions underlying the current debate about imprisonment policy are 1) Can the prison population be reduced in ways that do not increase the risk of crime and victimization to the public? and 2) Are prison sentences being meted out fairly and judiciously? Discussions about prison policy revolve around questions such as:

• Who gets punished with prison sentences? Is imprisonment being reserved for the most serious offenders, those that imposed the greatest harm on society or those that pose the greatest risk of recidivism, or are less risky offenders, those that could be punished by less costly community-based sanctions, also being imprisoned?

• Are the lengths of prison sentences appropriate? Do we get more safety and security for longer sentences and time served or do a few extra months in prison incur costs without generating public safety benefits through incapacitation or deterrence?

• What can be done to reduce the seemingly intractably high recidivism rates of prisoners? Do longer terms reduce recidivism or do they increase it?

• What are the collateral consequences of incarceration on society? What groups are disproportionately impacted, and if so, how much of that stems from sentencing and release practices?

• What are the implications of changes in the contribution of subgroups in the population for the costs of incarceration? As the prison population has aged, what are the implications for costs and care of elderly prisoners?

BJS statistics about changes in the size and composition of the prison population, changes in flows—admissions and releases—in length of stay, about returns to prison (parole recidivism) are central to understanding how state courts’ and parole supervising agencies’ decisions about who to send to prison reflect changes in practices. While it may be beyond BJS’ mission to address some of the evaluative issues that charge the debate about the scope and purpose of prison policy, BJS’ descriptive accounts of changes in the prison population set the terms of the debate. BJS uses the NCRP, in combination with other BJS corrections statistics, to provide this description. The distinctiveness of the NCRP (individual-level administrative records), however, makes it central to BJS in accomplishing its mission of describing changes in the corrections system.

To describe and explain changes in the size and composition of states’ prison populations, and the transition between incarceration, community corrections, and release into the general community, detailed individual-level administrative data are needed to augment the aggregate counts obtained in other BJS collections. The NCRP is the only national database that can routinely provide this individual level data. Changes in the age structure, racial/ethnic composition, or sentenced offense profile of prison admissions, prison releases, prison stock population, and the parole (that is, post-custody supervision) population can be described over time by NCRP in relation to other demographic and criminal characteristics that are impossible in aggregate counts of these populations. The NCRP data can be linked to external community data to better understand the re-entry of inmates and parolees through geographic variables collected. Since variations exist between and within states in the laws, implementation of laws, sentencing statutes, and community characteristics, data must be collected on the state level but also contain appropriate indicators for meaningful units of sub-state aggregation, such as the county of sentencing. Finally, to aggregate the state level data up to the national level or to produce national level estimates of important quantities, the data must be arrayed in a common format, in which the variables reliably measure core concepts across states.

The individual level data available in the NCRP are used alone and in combination with aggregate data on prisons and prison populations to inform a variety of important issues in correctional policy including:

1) Documenting the increase in the use of prison for violent offenders. Between 1999 and 2009, the proportion of state prisoners sentenced for a violent offense increased from 47.9% to 53.2% and the proportion held for drug offenses declined from 21.1% to 17.8%. These data suggest that prison space has been increasingly reserved for offenders who generated the most harm to society.

2) Showing how sentence lengths have changed over time and particularly documenting the shift from indeterminate to determinate sentence lengths and the relationship between changes in sentencing practices and sentence lengths.

3) Documenting how time served in prison has been slowing, both overall and within offense categories. During the current decade, for example, expected time to be served on a commitment has been relatively flat, with the mean time served moved between 2.0 and 2.1 years across the 10 years. The implication of this finding is that changes in prison admissions is the driving force behind changes in (and the recent decline in) the prison population.

4) Estimating parole recidivism. While this measure does not encompass all types of mal-behavior following criminal justice system contact, it is an important measure for state prisons as it generates expectations about future prison populations stemming from those released in a current period. And during the past two years (as described below in Part A, Section 2, Needs and Uses), BJS has invested in linking NCRP records to expand the scope of coverage of recidivism beyond only those prisoners released on parole. While this effort to enhance NCRP linkage is not yet complete, it represents BJS’ attention to key issues in the policy debate about imprisonment and BJS efforts to use low-cost methods to improve the scope of coverage of its statistics.

5) Generating and tracking changes in sex by race by age distributions of state prison populations and uses these to identify groups that may be disproportionately affected by prison.

6) Documenting declines in racial disparities in imprisonment, particularly among women. Among women, the ratio of the black-to-white imprisonment rates (number of prisoners per 100,000 residents) has declined from about 6 to 1 in 2000 to about 3 to 1 in 2010. The sources of this decline in racial disparity are declines in the number of black women in prison and increases in the number of white women in prison.

7) Showing that the age structure of the prison population has changed fairly dramatically over the past 20 years. The number of persons in prison who were 55 years or older more than doubled between 1990, when they represented 3.1% of the total prison population, and 2010, when they represented 7.9% of prisoners. Because the admissions and releases of inmates age 55 years or older grew at approximately the same rate during this period, BJS has demonstrated that it is the increase in violent offenses committed by older individuals, and their associated longer prison sentences, that is driving the increase in aged prisoners.

The Omnibus Crime Control and Safe Street Act of 1968 (see Appendix A), as amended (42 U.S.C. 3732) authorizes the Department of Justice, Bureau of Justice Statistics (BJS) to compile data on state and federal admissions to prison, releases from prison and discharges from community supervision (parole). Under Title 42 of the United States Code, Section 3789, BJS collects NCRP data for statistical purposes only, does not release data pertaining to specific individuals in the NCRP, and has in place procedures to guard against disclosure of personally identifiable information. NCRP data are maintained under the security provisions outlined in U.S. Department of Justice regulation 28 CFR §22.23, which can be reviewed at: http://bjs.ojp.usdoj.gov/content/pub/pdf/bjsmpc.pdf.

2. Needs and Uses

In 2009 at BJS’ request, the Committee on National Statistics (CNSTAT) of the National Academies issued its report detailing the findings of its review of BJS.¹ The CNSAT review addressed the full range of programs at BJS and examined (among other things) how BJS statistics are used in order to determine the impact of BJS programs and the means to enhance the impact. The CNSTAT report identified several gaps in coverage of BJS corrections statistics and several improvements to its portfolio of corrections statistics that could assist in filling the gaps. The gaps were: (a) expand coverage to include prisoner reentry and recidivism issues; (b) emphasize the flows and transitions of inmates in the corrections system; and (c) develop and enhance longitudinal datasets so that the data cover more steps in the criminal justice system instead of cross-sectional surveys on discrete parts of the system. BJS is using NCRP to address these three gaps.

Expand coverage to include prisoner reentry and recidivism issues

First, BJS uses the NCRP data as the base for its national studies of recidivism of released prisoners. Two of these have been completed and a third is in progress. The first was done for a prisoner release cohort in 1983 and the second for prisoners released in 1994. Currently BJS is completing data file preparation for the third, which will track five years of post-prison recidivism events (re-arrests, re-convictions, and re-incarcerations) of prisoners released in 2005 from 30 states. The NCRP data are used to draw the sample for the study and to provide information about inmates’ demographic attributes, offenses, sentences, time served, admission type, release methods, parole supervision and additional variables that may be associated with recidivism. BJS obtains data on recidivism events through the Federal Bureau of Investigation’s Interstate Identification Index (III), a fingerprint-supported computerized pointer-system for obtaining the record of arrest and prosecution (or RAP sheets) on arrested persons.

Due to the costs and complications associated with obtaining the data, BJS’ national studies of recidivism of released prisoners have been done periodically, at about 10-year intervals. To remedy this in part and obtain more frequent statistics on at least a portion of recidivism—re-incarceration—during the past two years, BJS has invested in improving linking of NCRP records across stages of the corrections process. Most notably for recidivism issues, working with its data collection agent, BJS has developed methods for identifying “terms” in prison, where a term is defined as the period from admission to release, and linking terms within commitments, where a commitment is defined as an entry to prison on a new sentence. An inmate may have more than one admission on a commitment if that inmate is released from an initial commitment but subsequently returns to prison because of a violation of conditions of supervision. White papers describing the concept and analysis of term records and their detailed construction are located in Appendices B and C.

BJS can use the NCRP to measure this portion of recidivism—return to prison from an initial commitment, or parole recidivism—and report on it routinely. This measure is valuable in the current debates on the use of prison. As mentioned above, in the debates about the use of prison, the argument is that certain classes of lower-risk offenders can be released from prison early and managed in the community for lower costs than those associated with prison, and supervised in ways that do not impact public safety. In states in which offenders are routinely released to community supervision, analysis of recidivism while under community supervision provides indications of the success or failure of community supervision. BJS can use the NCRP to inform this issue. Currently, BJS and its contractor have completed the analysis and linking of term records for 18 states, with an additional 26 states’ data still being processed.

Through the NCRP data collection agent, Abt Associates, Inc., BJS has conducted preliminary analysis and “proof of concept” work. In the states used in this analysis of linked data, repeated admissions were used to construct a hazard model to calculate the probability of recidivating given a number of demographic and temporal factors. Looking at the data over time, BJS and Abt were also able to determine that recidivism was declining in addition to identifying groups that were more likely to recidivate (see Appendix B for this study).

To further enhance the measure of return to prison, in this clearance, BJS is requesting clearance to add the FBI fingerprint number to the list of identification variables in the NCRP. BJS currently obtains the state ID (SID), which is a fingerprint number that is unique within states but not across states. The FBI compares state fingerprint records to identify unique fingerprints that are assigned FBI numbers. BJS uses the SID number in linking NCRP records to create the term records within states, but with this number, BJS cannot link across states. Hence, the value of the FBI number to the NCRP is threefold: (1) it will allow BJS to measure movements of inmates between states; (2) it will allow BJS to measure repeated entries into prison by the same person over time; and (3) in future national recidivism studies, BJS will have all of the information that it needs from the NCRP to obtain RAP sheets from the III system. With the second use of the FBI number, BJS will be able to develop life-course histories of recidivism as measured by imprisonment.

In addition, while BJS can use the term records to measure within-state recidivism, it needs the FBI number to measure out-of-state recidivism, as the SID numbers cannot be used for these purposes. Based on preliminary analysis that BJS has done using RAP sheet data in its current national study of recidivism of prisoners, BJS found that 15% to 55% of prisoners have been arrested in multiple states. Further, based on its efforts through the Survey of State Criminal History Information Systems (OMB control #1121-0312, BJS has learned that the since 2010, all fifty states make use of the FBI’s National Crime Information Center (NCIC), which allows states to search the Interstate Identification Index (III), a nationwide repository of fingerprint-supported criminal history records for persons arrested for felonies or serious misdemeanors. This system captures both SID number, as well as the unique FBI number. To participate in III, state information systems must have a criminal history record format compatible with the III system so that searches of the system from other states will result in standardized records; FBI number is captured by each state for these records. Obtaining this variable will permit BJS to create nationwide term records and describe cross-state crime and recidivism patterns. Since this variable is usually located in the same database as the demographic variables currently provided by states for NCRP, BJS anticipates the addition of FBI identification number to result in only a minor, one-time increase in burden (to revise the extract that creates the NCRP) while the benefits of adding this field to the NCRP are high, as indicated above.

Somewhat related to recidivism but important in its own right is mortality after release from prison. A recent study in Washington State has shown that former inmates have an adjusted risk of death that is 3.5 times higher than individuals in the general population.² Through its Annual Survey of Parole (OMB Control Number 1121-0064), BJS reported that the aggregate number of deaths on parole reached 5,700 in 2010, which puts the crude mortality rate on parole at nearly three times the crude mortality rate in prison (or 730 per 100,000 vs. 257 per 100,000). To better understand mortality post-release from prison and parole, BJS will link the NCRP to data from the National Center for Health Statistics (NCHS) National Death Index (NDI). The NDI is a repository of all deaths occurring in the U.S., and the NDI data include data on date of death as well as up to 10 cause-of-death codes.

This effort to link the NCRP to the NDI will both complement other BJS efforts to measure mortality in the criminal justice system (notably, the Deaths in Custody Reporting Program, OMB Control # 1121-0249), and provide information on outcomes during the corrections process. For example, based on preliminary analysis of NCRP data (which contain information on fact of death), deaths following release from prison are concentrated in the early months after release, and if Binswanger’s research in one jurisdiction holds across other jurisdictions, the majority of post-prison release deaths are preventable deaths, as they are caused by drug overdoses, homicides, and suicides. By comparison, in prison, the majority of inmate deaths are due to medical causes such as cancer and heart disease.

Finally, to improve estimates of recidivism, decedents need to be removed from the denominator of inmates at risk. Given the relatively high annual mortality rates on parole, failure to take mortality into account can affect recidivism estimates, particularly if post prison mortality and recidivism are correlated with risky behaviors that are associated with both outcomes. Former prisoners who return to substance use and abuse, for example, may also be more likely to engage in criminal activities, suggesting that the same causal factors may lead to both mortality and recidivism.

In sum, the NCRP data play a critical role in BJS’ efforts to understand recidivism.

Emphasize flows and transitions of inmates through the corrections system

The CNSTAT panel’s second major recommendation for BJS corrections statistics was to shift BJS’ focus towards emphasizing flows and transitions through the corrections process. To address this, in 2010 with the selection of Abt Associates as BJS data collection agent for the NCRP, BJS began to utilize the information in the NCRP to link records across NCRP cohorts to enable BJS to better understand movements into and out of prison and movements back into prison from parole. Through Abt, BJS undertook a major effort to construct “term records” (mentioned above) and “term histories” for state prison inmates (see Appendices B and C for detailed descriptions of term record creation). BJS used the state inmate ID variable from the NCRP records to link an inmate’s prison admission record (NCRP-1A) with one or more yearend stock record (NCRP-1D), and finally with a prison release record (NCRP-1B). A current limitation in the construction of term records is associated with the NCRP-C records, the parole discharge records. BJS is requesting clearance to add two variables to the NCRP-C records so that it can create parole terms and improve the linking of parole terms back to prison admissions. The two enhancement variables are date and type of admission onto parole. Parole refers to any type of post-prison supervision and type of admission onto parole refers to the manner by which an offender entered parole (or post-prison supervision) such as parole board decision, mandatory release onto parole, or expiration of sentence followed by supervision.

The term records inform the transition of prisoners throughout the imprisonment process and address the flow of prisoners over time at state and national levels. The term records yield several enhancements in measuring prison population changes. One important enhancement is the capacity to generate a stock population for any given day within a year. Prior to the construction of term records, only the yearend stock population was available. Because the term records associate an admission to a release, or if a release is pending, the yearend stock, BJS can use dates on term records to produce the prison population on any given day. Not only does this allow BJS to represent how prison populations fluctuate but also this allows BJS to identify seasonal variations that are related to prison population management decisions. For example, using its aggregate counts from the National Prisoners Statistics (NPS, OMB Control #1121-0102, BJS demonstrated a form of seasonal variation in the prison population in which the 6 months from January to June consistently showed an increase over the prior six months from July to December.³ But with the aggregate data BJS was not able to determine the source of the variation. With the NCRP term records, BJS can determine the source of the variation. If seasonal variation in prison populations is not related to seasonal variation in criminal justice system processing, but rather reflects choices made by prison officials to address, for example, overcrowding or other concerns, then analysts could take advantage of this variation to assess whether otherwise comparable prisoners released at different points throughout the year had comparable recidivism rates. Without the term record construction, the generation of daily prison populations, and the identification of the sources of the daily fluctuations, it would not be possible to capture this potentially important source of variation for the study of recidivism.

Another key transition that can be observed or estimated by the construction of the term records is the time spent in prison from admission to release. Time served from admission to release is of fundamental importance to understanding how prison populations grow and for understanding the impacts of sentencing reforms on prison populations. The concept of time served in prison—difference between release and admission date—seems to be deceptively simple but is actually complex to measure. Most commonly, time served is measured by those released from prison, that is, time served by a release cohort. This measure is useful for some purposes, such as assessing the impacts of time served on recidivism, but it is not useful for other purposes, such as assess the impacts of sentencing reforms on the severity of punishment or for use in forecasting prison populations. For example, if a sentencing reform is implemented at time T, time served by release cohorts would not give good measures of the impacts of the reform because the release cohorts consist of mixtures of admissions cohorts, many of whom entered prior to T. To assess the impacts of sentencing reforms on the severity of punishment, time served needs to be associated with the admissions cohorts at T and subsequent periods. Similarly, using time served by an exit cohort in prison population forecasting will yield biased estimates of the size of future populations unless the prison system is stable (that is, admissions rates are constant and time served is constant). Over the past 30 years, the prison system in the U.S. cannot have been characterized as a stable population.⁴ Rather for these latter purposes, time served needs to be estimated for persons admitted into prison.

The term records are essential for estimating “time to be served on a commitment,” but estimating time served on a commitment is complicated by the fact that NCRP data are available for a limited observation window and for persons still in prison at the end of the observation window, time to be served must be estimated. BJS has worked with Abt Associates in developing a methodological paper on estimating time served using the NCRP data, and one of the proceeds of the paper will be a methodology that can be applied to the NCRP data to generate reliable estimates of time served over time. The paper is currently in draft form although BJS plans to publish the final version of the paper around the end of 2012 or early 2013.⁵ Through this methodological enhancement, BJS will expand the utility of the NCRP to address key issues related to the severity of punishment in prison and the transition from admission to actual and expected releases. In sum, the longitudinal nature of term records makes it possible for BJS to observe the effects of new corrections policies on the prison population, as well as describe changes in the population demographics, sentencing, and offense characteristics over time. Only through the NCRP can BJS address these issues.

Develop and enhance longitudinal datasets to cover more steps in the criminal justice

The NCRP’s role in the third recommendation identified by the CNSTAT is limited, given that NCRP pertains to the prison system and not the entire criminal justice system. However, the NCRP has a role in BJS’ efforts to develop longitudinal datasets to cover more steps in the criminal justice system. BJS is making efforts to improve its court statistics and has only just started a new project to collect administrative data directly from state courts to generate its statistics on felony sentences in state courts. At this stage, BJS envisions that the NCRP data can be linked directly to the court data. If these efforts are successful (BJS has not yet obtained clearance to collect the court administrative records but plans to submit a request for clearance later this summer. BJS’ plan is to obtain linking variables in the court data that are comparable to the linking variables in the NCRP; for example, the SID or FBI number. If these are obtained, then BJS would link court records directly to the NCRP and also link NCRP prison release and parole discharge records back to court administrative records. If successful, this effort would allow BJS to estimate more precisely key quantities that it uses to measure transitions through the criminal justice system. These would include, for example:

• Admissions into prison directly from court sentences (versus other methods);

• Offenders initially sentenced to probation (a community based sanction) whose probation was violated resulting in a new prison admission (which BJS currently cannot measure precisely);

• Released prisoners who appear in felony court and their dispositions;

• Released prisoners who are on parole but who appear in court and are subsequently sentenced for a new crime.

While this is in the planning stages and the efficacy of these efforts will not be known until BJS obtains the court records, the planning reflects BJS’ efforts to expand its use of administrative records for statistical purposes.

BJS’ recurring uses of NCRP

In addition to using the NCRP to help to fill the gaps in BJS corrections statistics that were identified by the CNSTAT panel, BJS first and foremost uses the NCRP regularly to generate statistics that are key to understanding changes in the composition of prison populations. Primary among these are BJS efforts to accurately describe the age, race, sex, and offense distributions of state prison populations. Annually, BJS uses NCRP data in combination with other data from the National Prisoners Statistics to estimate these distributions. These estimates are published in the annual Prisoners bulletin.⁶ White this publication is based primarily on the data from the National Prisoner Statistics (NPS; OMB control # 1121-0102) collection, states only supply aggregate counts of admissions, releases, and stock populations to NPS; to describe the flow of state prisoners in terms of the offenses they have committed and their demographic profiles, BJS relies on the individual-level NCRP data. Also annually, using demographic methods of life expectancy that adjust for growth rates in the population, BJS uses NCRP to generate aggregate estimates of mean expected time to be served.

BJS uses the NCRP for special reports to address topical issues in corrections practice. For example, staff are currently working on a paper on the aging of the prison population that focuses on the offenses of older prisoners and the extent to which the upward shift in the age distribution of the prison population arises from changes in the age distribution of prisoners at admission or changes in time served by older prisoners.

Or, with increasing interest on the part of state-level stakeholders in comparing changes in their prison populations to other states, BJS is using the NCRP project to identify and document reporting differences among participating NCRP states and to prepare state profiles that characterize each state’s sentencing structure and prison release practices. These fact sheets are created from available published information, so in compiling them, there is no burden placed on state respondents. (An example of one state profile is included in Appendix D.) The fact sheets will serve three purposes: (1) As a special topic report on state sentencing practices, they will show the extreme variation across states in sentencing structure and release practices that point out difficulties associated with making simple comparisons between states on any measure related to prison populations; (2) Because the state profile information will also be coded into a database, they can be used to classify states that share common elements and facilitate making comparisons between states within groupings of states. (3) The information in them serves as a template for a future addendum to the NCRP or another collection (such as the NPS) so that the information can be updated on a routine basis and a longitudinal record of changes in practices will be created.

BJS is using the NCRP term records in a study of time served in prison that will result in the aforementioned methodological paper as well as substantive papers on time served by specific groups of prisoners. Other special topic reports using NCRP data that are planned but not yet underway include: factors affecting changes in racial disparities in prison populations and trends in state parole outcomes.

In response to requests for information from various sources, BJS statisticians use NCRP data to respond to questions from state and federal legislators, the press, and general public on issues related to corrections. In particular, NCRP data are used to address issues of trends in demographic and offense distributions over time. NCRP collects up to three convicted offenses from each inmate in addition to up to two offenses committed since the inmate was incarcerated. Variables describing multiple offenses and the sentencing information associated with these crimes, as well as the type of prison admission, provide a more complete picture of time served in prison, particularly in the calculation of percentage of an original prison sentence actually served.

Given the changes underway in California to reduce its prison population following the U.S. Supreme Court decision (Brown v. Plata, 2011), BJS is using the NCRP in concert with the National Prisoner Statistics (NPS), Annual Survey of Jails (ASJ), and Annual Surveys of Parole and Probation (ASPP) to capture data on California’s efforts to decrease its state prison population and to describe the changing prison population in that state. California is required to reduce its prison population by 46,000 inmates before June 27, 2013, which will decrease the operation of prisons at 200% design capacity in 2010 to 137.5%. California’s plan is to reduce the prison population through attrition while housing new nonviolent offenders in local county jails, but not to let existing prisoners out before their sentences are served, nor to transfer existing prisoners from prison to jail. Among other issues, BJS will use the NCRP data to determine (1) the rate of the California prison population reduction; (2) whether time served on sentences decreases compared to time served on sentences served in previous years for the same crimes; (3) the changing demographic and offense profiles of the California prison population as prisons house predominantly violent offenders in future years; and (4) the nature of California’s parole system, which used to provide the majority of prison admissions through parole revocations, but now will force parole violators to serve their time in jail unless their offense was violent. BJS will use NCRP variables on type of release from parole and supervision status at time of release to determine whether the movement of parolees back into the community has changed in size or method from previous years. As other states grapple with problems of overcrowding and smaller budgets, California’s efforts at controlling population size will be examined to determine whether such methods could be used in other jurisdictions. NCRP data will allow BJS to evaluate these efforts by looking at changes in the prison and parole populations before, during, and after the realignment of prisoners.

BJS continues to use the NCRP data to study racial and ethnic differences in the state prison population, as well as ways to improve collection of racial and ethnic data across all BJS surveys. At the request of OMB during the previous clearance process for NCRP (2009), BJS prepared a report addressing several practices of racial and ethnic data collection not compliant with OMB guidelines (this report was uploaded into the OMB system under ICR reference number 201112-1121-003). Chief among OMB’s concerns was the inclusion of both “other” and “unknown” race categories in the NCRP. Since that clearance process, BJS has replaced these categories in NCRP with “additional categories in your information system” and “not known”, and added a response option for “two or more races”. BJS’ data collection agent stressed the changes in phone calls to the data providers at the beginning of the collection phase, and while several states have responded that they can populate the “two or more races” category, most indicated that their administrative data systems are not structured to capture these data. BJS will continue to work with the states to encourage them to adopt the OMB classification system.

OMB was also concerned with BJS’ use of two questions to ascertain (1) the race; and (2) the ethnicity of each inmate for NCRP, and then completing the combination of the two questions during the data analysis phase. Many of BJS’ collections have been modified to obtain these data in a single-question format, reducing burden on our data providers. In the case of NCRP, however, a transition to a single-question format would require additional burden on our respondents, as they would need to rewrite their data extraction programs. BJS retained the two-question format in another individual-level administrative data collection, on capital punishment (OMB control # 1121-0030).

The BJS report on race submitted to OMB in 2011 compared NCRP racial and ethnic data to these fields in other collections, matching individuals whenever possible (example – decedents in state prison could be matched between the NCRP and Deaths in Custody Reporting Program (DCRP; OMB control # 1121-0249). BJS was especially interested to see whether the distributions of racial and ethnic groups differed between administrative data collections, like NCRP, and self-report surveys, including the National Inmate Survey (NIS, OMB control # 1121-0311). Differences between NCRP and NPS were minimal over a 10-year period (2000-2009), and when comparing NCRP data matched to individuals in the DCRP and capital punishment collections, the race/ethnicity values match for 99% of the cases where the inmate was White, non-Hispanic or Black, non-Hispanic, and 94% of the cases where the inmate was of Hispanic/Latino origin. Differences between NCRP and self-reported race/ethnicity in the NIS were larger, but this could be due to the fact that NIS does not sample from prison hospitals and halfway houses while NCRP data respondents are instructed to include inmates in these locations. Overall, the NCRP race/ethnicity distribution for 2008 was 4-5% higher for non-Hispanic Whites and Blacks than observed in NIS, and 2% lower for Hispanic inmates than reported on the NIS. BJS will continue to compare the race and ethnicity variables between NCRP and our other data collections to ensure that the mode of data collection (administrative versus self-report) does not influence the racial/ethnic distribution BJS uses for national estimates.

During 2011, BJS used NCRP data in presentations to the American Correctional Association and the University of Massachusetts Medical School Correctional Health Conference on the changing composition of the prison population in relation to the flow of inmates into prison and their length of stay once incarcerated. This introduced the data collection to a wider audience of practitioners and researchers, several of whom have expressed interest in using the archived NCRP data in their research.

BJS has used data from the NCRP in older reports to document rates of recidivism (see Recidivism of Prisoners Released in 1994, NCJ 193427), success rates among parolees over time (see Trends in State Parole, 1990-2000, NCJ 184735), changes in sentencing and time served (see Truth in Sentencing in State Prisons, NCJ 170032), changes in the number of drug offenders entering and exiting prison (see Drugs and Crime on the BJS website: http://www.ojp.usdoj.gov/bjs/drugs.htm), characteristics of female inmates (see Women Offenders, NCJ 175688), changes in the number of juvenile offenders in prison (see Profile of State Prisoners under Age 18, 1985-97, NCJ 176989).

Proposed enhancements to NCRP to expand its utility to address key issues

One of the most frequently asked questions of BJS is about the number of military veterans serving sentences in prison. Given news reports about veterans and crime, suicide, and other social problems, interest in criminal justice outcomes of veterans has grown. Currently, BJS captures information on veterans in prison only through is inmate surveys, the last of which was conducted in 2004 and the next one is not planned until 2013. To remedy this situation and to able to generate statistics about veterans in prison on routine basis, BJS plans to add to the list of NCRP variables three that capture veteran status information: ever served in the military (yes/no); date of last discharge; and type of last discharge. These variables comport with questions from BJS’ Survey of Inmates in State and Federal Correctional Facilities (OMB control #1121-0152) and the Department of Health and Human Services’ National Survey on Drug Use and Health (NSDUH; OMB control #0930-0110). Adding veteran status information to the NCRP will also bring it in line with BJS’ current effort to measure Problem Solving Courts, submitted to OMB for clearance on 6/27/2012. This collection will capture administrative details of state courts that have developed special dockets to handle defendants who have ongoing social and/or psychological conditions that underlie their repeated contact with the criminal justice system; recently, this has included some military veterans. Consistent with BJS plans to link NCRP with court records, capturing veteran status in the NCRP and comparing estimates with information from the Problem Solving Courts will provide for consistency in the measures of the volume of veterans in the two systems.

Based on prior BJS efforts and discussions with stakeholders from the Veterans Administration (VA) who work with departments of corrections, BJS has concluded that these measures are generally collected at intake as reported by the inmate and recording in information systems. In 1998, BJS conducted an inventory of state corrections information systems. At that time, 45 departments of corrections could identify whether a prisoner had served in the U.S. military, and 41 could give information on the inmate’s last military discharge. BJS contacted Dr. James McGuire, the National Director of the Veterans Health Administration’s Justice Programs, to check that the items BJS intends to collect (ever served in the military, date of last discharge, and type of last discharge) would be sufficient to describe an incarcerated veteran’s military service. Dr. McGuire said that while these variables might be at risk for underreporting by veterans, he agreed that these variables would be a valuable addition to the NCRP. More recently, BJS had conversations with NCRP data providers in 9 states regarding whether they would be able to provide data on veteran status, FBI number, and the proposed new parole variables, and all have indicated that they would be able to do so without any problem. Since the veteran status variables are usually located in the same database as the demographic variables currently provided for NCRP, BJS does not anticipate the addition of these items to result in a significant additional burden to the state data providers beyond the initial reporting year’s change in programming to request these new variables. BJS is adding one hour to its estimate of time needed to modify the extraction programs for each NCRP file type to include military status variables.

The second enhancement proposed is to the NCRP-1C, parole records. The NCRP-1C is meant to augment the aggregate statistics collected in BJS’ existing annual parole survey in the same manner as the prison admission, release, and yearend census records do for NPS: provide a more complete description of the parole population through the analysis of individual-level data. Over the years, BJS has recognized that the parole records are the least reliable of the NCRP record types, in large part because six states separate their parole boards from their departments of corrections, and do not share common data systems. To improve the reliability parole records, and enhance the prison term records by linking community corrections details to describe the entire sentence of inmates, BJS is proposing to add date and type of parole admission variables, as well as the location of the parole release or the parole office for each individual starting with the 2012 reporting year. Justification for collecting these variables is as follows:

• Date of admission to parole/post- confinement community supervision - Because six states have separate governmental respondents providing the prison and parole data, a prison release record (NCRP-1B) indicating that the inmate was released to parole/post- confinement community supervision does not always have a corresponding parole release record. Patterson (2010) found that some states provided up to 20% fewer parole release records to NCRP than would be expected from other BJS data sources.⁷ BJS performed an analysis of states’ ability to link prison release to parole release records in 2011 through the NCRP data collection agent, and found that while a number of states achieved matching in over 90% of the cases, those states with separate departments of corrections and parole boards had very low matching rates (below 50% in some cases). In addition, the initial design of NCRP assumed that inmates released from prison to parole would begin their post-custody sentence immediately upon exit from incarceration, so the date of parole “admission” was not included in the data elements requested for NCRP-1C. Adding this item will permit a more accurate matching of prison release records to parole counterpart records by giving BJS an additional limiting factor (date of prison release compared to date of parole entry) on which to make links between records. The addition of parole admission date will also provide a more accurate estimate of time served in community corrections. Since this variable and those following are located in the same database as the date of parole release currently provided for NCRP, BJS does not anticipate this new request to result in any additional burden to the state data providers beyond the initial reporting year’s change in programming to request this new variable. However, to account for this initial reporting year addition/modification of 4 parole variables, BJS has added an additional hour of programming to the burden estimate for parole record extraction.

• Type of admission to parole/post-confinement community supervision – As with the date of admission to parole/post-custody community corrections, the type of admission will assist BJS in linking prison release to parole records. Since this variable is located in the same database as the date and type of parole release currently provided for NCRP, BJS does not anticipate this new request to result in any additional burden to the state data providers beyond the additional hour accounted for in the date of admission to parole variable description.

• County of parole release/location of parole office – NCRP already collects the county in which the original sentence was imposed on the NCRP-1A, NCRP-1B, and NCRP-1D records; adding the county in which the parolee was released or the location where the parolee reported to supervision authorities will allow BJS to better understand the process of reintegration of former prisoners into society. Specifically, BJS would be able to determine whether parolees tended to stay in the same community as where they were sentenced, or if they moved to a different county or state. Combined with the FBI identification number, this variable would let BJS analyze recidivism by parolees across states. Since this variable is located in the same database as the date and type of parole release currently provided for NCRP, BJS does not anticipate this new request to result in any additional burden to the state data providers beyond the additional hour accounted for in the date of admission to parole variable description.

• Update of supervision status at time of release responses – To bring the NCRP in line with BJS’ annual parole survey, BJS is requesting the addition of one categorical response to the current question on NCRP-1C that asks for a parolee’s status at the time of his/her release from parole. Since the design of the NCRP-1C, a number of states have introduced a type of supervision that only requires the parolee complete certain financial obligations instead of active or inactive supervision by a parole agent. BJS obtains aggregate data on individuals under this type of supervision in our annual parole survey; adding the category of “only have financial conditions remaining” to NCRP-1C would make it possible for BJS to describe the types of offenses or demographic characteristics of parolees associated with this type of parole supervision versus the others. Since this is an additional response to a currently requested data element in NCRP, BJS does not anticipate this will result in any additional burden to the state data providers beyond the additional hour accounted for in the date of admission to parole variable description.

• Change in scope of records collected – Several states have notified BJS that they are unable to submit parole data because their data files on parole/post-confinement community supervision are not organized in such a way as to let them select only those individuals released from parole during a given year. To increase the number of states submitting parole data, BJS will ask these states to submit an annual “touch” file, which would collect the same variables request in the NCRP-1C form, but would include any individual who entered, exited, or was on parole during the year and would encompass the entire parole population for the state. This would reduce the burden on the states to provide the data to BJS, although it would increase the burden on BJS’ data collection agent to process these records.

External research uses of NCRP

BJS makes the NCRP data set available to the public through public use files located at the National Archive of Criminal Justice Data Archive at the University of Michigan. Between 2007 and 2011, NCRP data were downloaded by the public from the archives 5,785 times, with 1,753 of these instances in 2011 alone. NCRP data have been cited in more than 30 published reports or journal articles from 1999 to the present. Through 2009, BJS published detailed tables using NCRP on the BJS website (http://www.bjs.gov/index.cfm?ty=pbdetail&iid=2065) , in the future, BJS plans on making the NCRP data even easier to access through an online data tool that will permit the public to do cross-tabulations of age, race, sex, offense, and geographic location by year.

Annual estimates of race/ethnicity distribution of state prisoners based on NCRP data are routinely used by non-governmental policy advocates such as the Sentencing Project and Human Rights Watch to argue for changes to sentencing and incarceration practices, particularly in relation to specific offense types. Human Rights Watch used NCRP data to show differences in the rate of prison admissions for non-Hispanic White and Black individuals for drug offenses during the 1990s and 2000s, in concert with other BJS data demonstrating a racial disparity between Blacks and Whites for the ratio of percent population to arrests for drug offenses.⁸ Wildeman⁹ utilized the NCRP year-end population counts and racial distributions to examine the risk of inequality between children of incarcerated and non-incarcerated parents, and the disproportionate effect this has on non-White populations.

Dominating the current debate on prison policy are concerns about the size of U.S. prison populations and its effects on crime. Senator James Webb (D-VA) has introduced legislation in each year since 2009 that would require an extensive review of the criminal justice system with a focus on incarceration policies (the National Criminal Justice Commission Acts of 2009/2010 (S.714, H.R.5143) and 2011 (S.306)). Understanding the relationship between the decline in crime during the 1990s and 2000s and the growth of the prison population has been addressed by researchers and policy advocates using NCRP data. Some research has pointed out that prison population growth reflected policy choices and practices more than changes in crime rates. For example, Blumstein and Beck show that increased use of technical violations of conditions of parole contributed significantly to prison population growth during the late 1990s,¹⁰ and Raphael argues that sentencing policy is of first order importance in explaining the growth in the prison population.¹¹ And, with the increase in the U.S. prison population, some have also questioned whether additional increases in the prison population will be cost-effective in terms of crime reduction.¹² All of these studies relied heavily upon NCRP data.

John Pfaff used the NCRP data to analyze the role of changes in the amount of time served in prison in the growth of the prison population over time.¹³ Using data from 11 states that had consistently contributed data to NCRP between 1983 and 2002, he found that the amount of time served in prison has not changed appreciably over this time period (and in many states, has actually declined), despite legislative modifications of sentencing practices, and that median time served is relatively short, between 1-2 years. He concludes that increased admissions, particularly of minor offenses that did not result in incarceration prior to 1983, as well of parole violators, accounted for the increase in the prison population through 2002. He also points out that changing the admission rate of inmates with shorter sentences would not necessarily cause a perceptible decline in the total prison population, since those inmates with longer sentences could severely skew the distribution of population size. These results could affect the sentencing policies adopted by states in the future, since Pfaff argues that neither mandatory minimum sentences, nor habitual offender laws, are contributing as much to the growth in the prison population as the admission of individuals on short sentences who otherwise would not have been incarcerated.

Policy advocates for smaller prison populations also have used BJS data derived from the NCRP. Human Rights Watch published a study of the aging prison population using NCRP data from 1995 to 2009, noting that while the total state and federal prison populations grew 42.1% over this period, prisoners 55 years or older grew 282% as a group.¹⁴ They cite increased admissions of older inmates, longer sentences associated with more violent crimes, and mandatory minimum and habitual offender sentencing statutes as potentially contributing to the growth on older inmates. The study argues that older offenders are more costly to manage and keep healthy, have a greater risk of being victimized by younger inmates, and have lower recidivism rates, and states should consider passing laws to specifically address the needs and management of older prisoners.

The National Council on Crime and Delinquency (NCCD) examined the offense distribution from the 2003 NCRP to advocate for alternative measures of punishment for less serious crimes.¹⁵ Using the NCRP numbers from California, Florida, New York, and Texas, NCCD estimates potential cost savings for the states if less serious crimes were dealt with using electronic monitoring, drug courts, drug treatment, and community reporting programs like work release.

NCRP data have been used by external researchers to examine disparate subpopulations of, and policies affecting the prison population, including how alcohol control policies are related to violence;¹⁶ the possible effect felon disenfranchisement has had on past national elections;¹⁷ the discrepancy in sentences between male and female sex offenders;¹⁸ and modeling of the prevalence of HIV among released prisoners over time.¹⁹ The Pew Center on the States is currently using the publicly available NCRP data files to calculate time served in state prison for various offenses, and the National Bureau of Economic Research²⁰ employed drug-related incarceration data from NCRP as one measure of drug use among women during the 1990s welfare reform effort.

In addition, NCRP data have been used to address issues related to prisoner reentry;²¹ the effects of drug control policy on prison populations and black Americans;²² studies of economic models of crime applied to the criminal justice system in which NCRP data were featured prominently in studies of the prison population;²³ estimates of time served in prison and its relationship to sentencing policy;²⁴ and the NCRP data have been used to study the unintended consequences of incarceration, such as their possible effects on family formation.²⁵

In sum, the NCRP provides data on key national and state level issues in sentencing and correctional policies. Because the data are continuous and comparable among the states, the NCRP allows users to address such issues as current trends in sentencing and time served in prison and on parole, recidivism, rates of success for parole supervision and other key re-entry topics, the impact of mandatory sentencing practices on prison admissions and releases, the adoption of truth in sentencing policies and practices, the changing offense composition of inmates entering, exiting and under custody of the nation’s prisons, and the shifting demographic profile of inmates. If the NCRP were discontinued, decision makers would lose a valuable source of information for criminal justice system planning and policy formulation.

3. Use of Technology

The NCRP data collection uses standardized reporting items and formats. Most participating states provide data on an annual basis, allowing them to use existing computer extraction programs with very minor alterations. Through funding and technical assistance, BJS staff has worked with states to develop these programs, as well as update them when the states migrate to new information technology (IT) systems or database management software.

Data are extracted from individual states’ correctional databases, encrypted, and uploaded to the Abt Associates’ dedicated NCRP server using a password-protected secure File Transfer Protocol (FTP). Secure FTP was first implemented in January 2009 for use in collecting the 2008 NCRP data files, and is currently used by all of the NCRP respondents.

BJS provides the respondents with technical assistance as needed to minimize respondents’ efforts in data collection and to improve data quality control. For example, BJS staff recode state statutes and other offense codes to standard BJS codes, which significantly reduces the burden on participating jurisdictions. Quality control protocols implemented by BJS provide quick identification of out-of-range data values, abnormally high rates of missing data, and compare the current year’s data to previous years’ (if available) to ensure that large changes in the variable-specific and total number of data submitted can be explained by the state data providers. Rapid processing and quality control of the data results in reduced burden for states since any discrepancies can be immediately addressed.

4. Efforts to Identify Duplication

The National Corrections Reporting Program (NCRP) is not duplicated by any other federal government agency or program. No other program employs uniform criteria and comparable definitions when collecting data of this kind from state corrections agencies. BJS is the only government agency that collects national level data on sentencing, time served in prison and on parole, and offense composition for prison admissions, releases, and offenders in the custody of State prisons at yearend.

The Association of State Correctional Administrators (ASCA) conducts a state prison data collection called the Performance Based Measures System (PBMS; http://www.asca.net/system/assets/attachments/3460/**PBMS%20Resource%20Manual_August%202011.pdf?1315423025). This collection requires states to submit data monthly, requests 135 core elements over nine correctional areas, including mental health, budget, personnel, institutional security, and academic education. PBMS obtains some of the variables also collected by NCRP, including prisoner demographics, sentence lengths, and offense distribution, but these are collected only in the aggregate, not on an individual level as in NCRP. The purpose of PBMS, to provide corrections administrators with performance indicators over a period of time to allow for monitoring of the prison system, is very different from NCRP in form and content. BJS uses the NCRP administrative records to describe changes in state correctional populations and measure transitions and outcomes. The PBMS is an aggregate reporting system for state correctional administrators. Currently, only 14 states currently provide PBMS with at least half of the core elements on a monthly basis. BJS is collaborating with ASCA to see where NCRP can be used to supplement PBMS measures.

Similarly, BJS’ National Prisoner Statistics (NPS) collection obtains data on state prison admissions, releases, and yearend stock populations, along with race/ethnicity and sex, but these are aggregate counts. These totals are reported by BJS on a yearly basis to describe the state prison population and can serve as control totals for weighting, but more detailed disaggregation of the population, such as racial differences in time served on parole or changes in offense distribution by age groups over time, requires the individual-level data of the NCRP.

In the past, individual researchers have requested datasets encompassing similar elements to NCRP from a subset of state departments of corrections, primarily to obtain variables not collected in NCRP, or to get PII data that are not available to the public through the archived NCRP datasets. If the data are collected in NCRP, BJS encourages states to refer researchers to NACJD so that the state is not duplicating efforts in data provision.

5. Impact on Small Business

Not applicable. The NCRP data collection does not involve small businesses or other small entities. The respondents are state departments of corrections.

6. Consequences of Less Frequent Collection

Given the enhancements achieved by creating term records, less than annual collection of NCRP would greatly limit BJS capability to measure changes in the prison population, assess recidivism and reentry issues, enhance linkage of records to expand coverage of key issues related to prison populations, and measure transitions between stages of the correctional system. In addition, all annually reported measures of change in sex, race, age, and offense (and cross-classifications of these groups) composition of prison populations would be put at risk, as BJS would have to rely on older data and some type of weighting or estimation routine that might be insensitive to changes within any combinations of groups for which BJS produces these estimates. For example, in recent years, the percent of blacks incarcerated for drug crimes has dropped from 26.2% in 1999 to 21.1% in 2009. Less than annual collection of the data would prevent BJS from identifying these important changes.

The NCRP data are collected annually, and BJS has devoted resources to help expedite processing of the data in order to enhance its timeliness. Less frequent collection of data would impose burdens on respondents who have set up computerized methods to generate the NCRP data. In some states, data on admissions into and releases from prison are updated in real time, and if NCRP extracts of annual admissions and releases to be requested on a schedule other than annual, some states would only be able to provide records for those prisoners being admitted or released for the most recent twelve months. In addition, as NCRP is the only national data set that contains comparative data for monitoring trends in sentence length and time served, delaying or collecting data less frequently would impact BJS’ and other researchers’ ability to detect changes in sentencing practice that affect prison populations.

7. Special Circumstances Influencing Collection

Not applicable. The NCRP collection is consistent with the guidelines in 5 CFR 1320.6

8. Federal Register Publication and Outside Consultation

The NCRP collection is consistent with the guidelines in 5 CFR 1320.6. The 60 and 30-day notices for public commentary have been published in the Federal Register, no comments were received.

BJS maintains frequent contact with data providers and data users in an effort to improve data collection, reporting procedures, data analysis, and data presentation. In the case of NCRP specifically, BJS held a data providers’ meeting in 2011 and has scheduled a second meeting in October 2012 to discuss issues such as instructions for data submission, reporting format, item content, publication, archiving of data, and plans for web tool development.

In preparation for requesting clearance to add items on veteran status and past military service, BJS consulted with Dr. James McGuire, the National Director of the Veterans Health Administration’s Justice Programs. He provided valuable advice on the reliability of self-reported veteran data and made suggestions for the items to collect that would allow BJS to estimate the number of prisoners self-identifying as having prior military service. Dr. Evelyn Patterson of Vanderbilt University gave suggestions for improving the quality and reliability of the NCRP parole records, including the addition of date of admission onto parole.

9. Payment or Gift to Respondents

Not applicable. No payments or gifts are offered to NCRP respondents.

10. Assurance of Confidentiality

Under Title 42 of the United States Code, Section 3789, BJS collects NCRP data for statistical purposes only, does not release data pertaining to specific individuals in the NCRP, and has in place procedures to guard against disclosure of personally identifiable information. NCRP data are maintained under the security provisions outlined in U.S. Department of Justice regulation 28 CFR §22.23, which can be reviewed at: http://bjs.ojp.usdoj.gov/content/pub/pdf/bjsmpc.pdf.

11. Justification for Sensitive Questions

Not applicable. The NCRP questionnaire does not contain any sensitive questions.

12. Estimate of Hour Burden

There are 57 respondents in the NCRP data collection universe including the department of corrections (DOC) in each of the 50 states, the California Division of Juvenile Justice, and 6 different contacts for parole data in those states (Georgia, Massachusetts, Nevada, New York, Pennsylvania, and South Carolina) where the DOC does not keep data on parolees. The following burden estimates are calculated for two types of data providers: states that are currently providing NCRP data to BJS, and states that have either never provided data, or that have had a lapse in submission for a number of years. For the non-participating states, BJS identifies 24 key data elements across the 4 NCRP file types for the respondents to submit during the first year, then asks these respondents to submit the full complement of requested variables starting in the second year of participation, once they have extract programs created. Data on federal prison inmates are obtained through BJS’ Federal Justice Statistics Program (FJSP), so no burden is placed on the Bureau of Prisons for NCRP; parole data for the District of Columbia is likewise obtained through a different BJS data collection.

The costs to respondents incurred as a result of participating in this data collection are costs that would be incurred in the normal course of daily operations, except for the hours involved in preparing the data. Initial participation requires the one-time development of a computer program to extract data to be sent to NCRP. This computer program is then re-run to prepare data for submission in subsequent years. From discussions with both current and potential contributors to the NCRP, BJS estimates the time needed to develop computer programs to extract data and to prepare a response to be 24 hours, per type of database containing the information needed incurred during the first year of participation, and 8 hours per type of database during the second and subsequent years of data provision (prior to the addition of requested new items).

At the beginning of each year, states are contacted by BJS’ data collection agent for a brief phone conversation to confirm that the data respondent has not changed over the past year, and to tell them to expect a packet of materials describing the submission of NCRP data in the next few weeks (see Appendices E and F for scripts for the calls to new/lapsed and currently contributing states). In 2013, the packet for collection of the 2012 NCRP data will contain: an introductory letter from BJS (Appendix G); an introductory letter from the data collection agent (Appendix H); a newsletter of NCRP-related events from the past year (Appendix I); and depending on whether the state had submitted data in 2011, instructions for currently contributing or new/lapsed data contributors (Appendices J and K).

Hour burden for proposed new variables in NCRP

BJS estimates that there will be three additional burden hours per reporting entity due to the addition of needed variables and the review of state criminal justice fact sheets (for more detailed justifications of the need for these changes, please see the Needs section of this supporting statement):

New Variable	Justification for requesting variable	Additional burden estimate
FBI identification number	Will improve BJS’ ability to measure recidivism between states	1 hour (total)
Ever a veteran (yes/no)	Will allow BJS to give more timely estimate of veterans in state prison
Veteran – date of last discharge	Will permit BJS to determine military service era of inmate
Veteran – type of last discharge	Will allow BJS to place current incarceration in context
Date of parole admission	Will allow more accurate calculations of time served on parole; permit BJS to match parole records to prison records	1 hour (total)
Type of parole admission	Will make it possible for BJS to match parole records to prison records
County of parole release/parole office	Will provide information for re-entry studies
Update to existing parole status variable	New type of parole supervision; will bring NCRP data in line with annual BJS parole survey
Change in scope of parole records requested	Will allow states unable to provide only parole release data with a way to submit parole data through a file of all parole “touches”
Review of state criminal justice fact sheets	Will promote responsible data comparisons between states, since this provides a standard template for states’ laws, statutes, and institutions	1 hour

Burden hours for prison records (NCRP-1A, NCRP-1B, NCRP-1D)

The development of computer programs during the first year of submission to extract prison records with information on prison admissions (NCRP-1A) and prison releases (NCRP 1-B) is estimated to require a total of 24 hours, as these represent snapshots from the same database. The first year provision of information on persons in prison at year-end (NCRP-1D) is estimated to require an additional 24 hours, as these records reside in a separate database.

During all subsequent years of participation, the average time needed to provide data is expected to be 8 hours per respondent for prisoner admissions and releases (NCRP-1A and NCRP-1B) and 8 hours for data on persons in prison at year-end (NCRP-1D), based on conversations with data providers during follow-up calls. The average of 8 hours per database takes into account that some respondents just need 2 hours to make a copy of a research database, while others may need to do additional work, including modifying computer programs, preparing input data, and documenting the record layout. For 2013, this burden estimate is increased by one hour (to 9 hours for prison admissions and releases and 9 hours for year-end stock populations) from the previous NCRP OMB submission to include modifications to existing data extraction programs needed for military status variables.

Burden hours for parole records (NCRP-1C)

To address issues with the quality of the parole records, BJS reviewed and refined the NCRP-1C data request to be made in 2013. In addition to adding the date and type of parole admission/entry and the location of parole release or parole reporting office, BJS is requesting that states not currently submitting parole records due to the difficulty in selecting only parole release records instead submit a “touch” file which would encompass the same variables, but for all individuals involved in the parole system in that year, regardless of whether they were released or not.

BJS anticipates that these changes to the parole records will result in an increase of one hour in the burden for state respondents. Since the additional requested items (date of parole entry, type of parole entry, county of parole release or location of parole office) will be in the same database currently used to obtain parole release dates, their collection will not require additional work beyond updating the extract program. Therefore, we estimate a total of 10 hours of burden for currently participating states to produce 2012 data for NCRP-1C (one additional hour for military status and FBI identification variables, and one additional hour for changes to the parole variables).

Burden hours for follow-up consultations

Follow-up consultations with respondents are usually necessary while processing the data to obtain further information regarding the definition, completeness and accuracy of their report. These consultations, usually initiated by email and then followed up with a phone conversation if there is difficulty in understanding the request, vary depending on the data elements that require attention. A sample of 8 states’ follow-up consultations from the 2011 data reporting year is provided in Appendix L to demonstrate the range of issues addressed. Respondents are also asked to review their data after the processing is completed. This consultation, review and feedback are estimated to take 2 hours per respondent.

BJS has begun to evaluate the ability of NCRP to provide accurate between-state comparisons of prison and parole populations in addition to national estimates, and the first step in this process is to develop standardized fact sheets on the laws, statutes, and institutions in each state’s criminal justice system. BJS’ data collection agent for NCRP has researched and written these fact sheets, and proposes to request the states’ NCRP respondents to review their fact sheet before posting them to the NCRP data providers’ web site. For report year 2012, BJS is adding 1 hour of additional time to the follow-up consultation for each state. This additional hour will not be required after the initial year in which states review and approve their fact sheets.

In contrast with the 2009 reporting year when several respondents continued to provide data on physical magnetic media, all of the respondents now provide NCRP data through a secure file transfer protocol that allows for fast delivery and an elimination of mailing costs.

Total burden hours for submitting NCRP data

BJS anticipates 39 states (40 respondents) currently providing NCRP data to provide NCRP-1A, NCRP-1B, and NCRP-1D data for report year 2012, and 36 states (36 respondents) to submit parole records (NCRP-1C). Including the time needed to review their data submissions, the total respondent burden for report year 2012 among current NCRP participants is expected to be 1,200 hours, as depicted in the table below. This is equivalent to roughly 30 hours per respondent currently submitting.

BJS plans to make a concerted effort to obtain the 2012 NCRP-1A, NCRP-1B, and NCRP-1 D data from the 11 states not currently submitting and the NCRP-1C data from the outstanding 14 states. The total burden estimate for these newly submitting states is 930 hours.

The total burden for provision of NCRP data across the 50 states (51 respondents for each individual part of the NCRP) we anticipate will participate in 2013 (for report year 2012) is 2,121 hours. Once all respondents are submitting data and have incorporated the new variables into their data extract programs, the estimated burden will be 1,326 hours annually, or an average of 26 hours per respondent.

Estimated time burden for states submitting NCRP data in 2013 (report year 2012)

SUBMISSIONS	FILES					TOTAL
	NCRP-1A	NCRP-1B	NCRP-1C	NCRP-1D	Data review1,2
States currently submitting to NCRP						1,200 hours
Number of states	39	39	36	39	39
Total number of respondents3	40	40	36	40	40
DOC respondents	40	40	31	40	40
Parole respondents	0	0	5	0	5
Estimated burden/response	9 hours		10 hours	9 hours	3 hours
Subtotal	360 hours		360 hours	360 hours	120 hours
States not currently submitting to NCRP						930 hours
Number of states	11	11	14	11	14
Total number of respondents3	11	11	15	11	15
DOC respondents	11	11	14	11	14
Parole respondents	0	0	1	0	1
Estimated burden/response	24 hours		24 hours	24 hours	3 hours
Subtotal	264 hours		360 hours	264 hours	42 hours
Total submissions and burden in 2013						2,121 hours4
Number of states	50	50	50	50	50
Total number of respondents3	51	51	51	51	57
DOC respondents	51	51	45	51	51
Parole respondents	0	0	6	0	6
Estimated total burden	624 hours		720 hours	624 hours	153 hours

¹The estimated burden for data review is still assumed to be 3 total hours per NCRP submission of 4 parts, even though two different people in these six states will separately review the parole records (1 part) and the prison records (3 parts).

²The number of states, respondents, and burden hours in the data review column subtotals will not sum to the total submissions and burden hours, since these values reflect the maximum number of states and respondents that will need to review at least one submitted file of NCRP.

³While there are 57 total respondents for NCRP, only 51 respond to each individual part of the collection. In six states, a different respondent reports data for NCRP-1C (parole release records) than those states’ respondents for the other parts of NCRP. Currently, five of the six states submit data on an annual basis to NCRP.

⁴The estimated total burden for all 2013 submissions will not be the sum of the estimated subtotals of states currently and not currently submitting NCRP records because of the data review estimate issue described above.

Estimated maximum annual time burden for NCRP data submission, assuming all states participate (post 2013)

SUBMISSIONS	FILES					TOTAL
	NCRP-1A	NCRP-1B	NCRP-1C	NCRP-1D	Data review1
Number of states	50	50	50	50	50	1,326 hours
Total number of respondents2	51	51	51	51	57
Estimated burden/response	8 hours		8 hours	8 hours	2 hours
Subtotal	408 hours		408 hours	408 hours	102 hours

¹The estimated burden for data review is still assumed to be 2 total hours per NCRP submission of 4 parts, even though two different people in these six states will separately review the parole records (1 part) and the prison records (3 parts).

²While there are 57 total respondents for NCRP, only 51 respond to each portion of the survey. In six states, a different respondent reports data for NCRP-1C (parole release records) than those states’ respondents for the other parts of NCRP.

The total burden estimate of 1,200 hours for respondents currently submitting NCRP data has increased from the 2009 estimate of 842 hours due to the increase in number of respondents submitting NCRP-1C (9 additional respondents) and NCRP-1D (14 additional respondents). As previously stated, the burden estimates for all four types of NCRP records have been increased by one hour for the 2012 data submission year to accommodate extra time needed to program the extraction of military status variables, and an additional hour has been added to the NCRP-1C file to allow for the additional extraction of parole admission variables. An additional hour has been added to the data review for states to verify their state criminal justice fact sheet.

The burden of 930 hours for new NCRP submissions for report year 2012 (states not submitting 2011 or earlier data) represents a decrease from the 2009 estimated burden of 1,412, due almost exclusively to the increased participation in NCRP over the past three years.

Once full participation is achieved in NCRP, involving the provision of each NCRP file type (NCRP-1A, NCRP-1B, NCRP-1C, and NCRP-1D) by all respondents, the total annual burden is expected to be 1,326, a decrease of 928 hours from the 2009 estimate submitted to OMB.

13. Estimate of Cost Burden

The costs to respondents incurred as a result of participating in this data collection are costs that would be incurred in the normal course of daily operations.

14. Estimated Cost to Federal Government

The estimated costs for collection, processing, and dissemination of the NCRP data in 2012 is $933,120, including:

$800,000 -- Abt Associates, Inc.

$531,480 for data collection, data processing, and program management

$200,000 for computer programming, providing data, furnishing publication-ready tables

$68,520 in miscellaneous charges -- costs related to postage, telephone calls, disks to respondents, printing, etc.

$133,120 -- Bureau of Justice Statistics

60% GS-13, Statistician ($66,000)

10% GS-15, Supervisory Statistician ($14,000)

Fringe benefits (@28% of salaries -- $22,400)

Other administrative costs (@30% of salary & fringe $30,720)

15. Reasons for Change in Burden

The estimate of 2,121 hours for states to compile and submit the four parts of the NCRP in report year 2012 has decreased since the collection’s previous OMB clearance due to the number of new states submitting one or more data files in the past three years, as well as the adoption by all states of secure data transfer protocols without the involvement of time and cost spent on magnetic media preparation.

16. Plans for Publication

BJS’ plans for products and publications from NCRP data over the next three years fall into four categories and include the following (a calendar of proposed publications is located at the conclusion of this section):

Technical and methodological documentation

BJS and its data collection agent for NCRP will document the methodologies that is has used to create the term records, link NCRP records, and weight NCRP records to improve the transparency related to how it generates estimates from the NCRP. These documents will describe the details of the processes and procedures. These documents will be made available to the public for those interested in the development of the term records, but there are no current plans for formal publication of these reports. Four reports discussing the methodology of the term records are planned:

• Technical document #1: Observations on the NCRP – This paper details the idea of constructing term records from the prison admission, prison release, and prison yearend census records, and gives a brief overview of the methods to be used in the construction.

• Technical document #2: NCRP reporting (in draft format in Appendix B) – This paper uses constructed NCRP term records from four states to demonstrate the increased analytic capabilities of the term records. The probability and hazard of returning to prison as well as time served are calculated and discussed.

• Technical document #3: A description of computing code used to identify correctional terms and histories (in draft format in Appendix C) – The white paper is a methodological description of how term records are constructed, including the observed types of “ambiguous” term sequences (the sequence of prison admission, yearend census, and release records is not logical) in the NCRP and the decisions made in resolving these problems.

• Technical document #4: Adjusting for unit nonresponse in the NCRP – This paper discusses the statistical methods for adjusting the prison admission, yearend, and release NCRP data for states that do not contribute data in particular years, so that BJS can create national estimates of detailed criminal and demographic characteristics not available in the National Prisoner Statistics data collection.

BJS Technical Reports

Currently one technical report is planned:

• Estimating Prison Time-Served with Piecewise Constant Hazard Models: Problems and Applications using National Corrections Reporting Program Data. This paper will fall into BJS’ “technical/methodological” report series which is designed to showcase new methodologies that apply to a BJS statistical program or series. The paper will describe the methodology, assess it relative to Monte Carlo simulations, and illustrate the use of the methodology in estimating time served for several population groups. The planned publication date is late 2012/early 2013.

BJS Bulletins

BJS bulletins provide the “first cut” from a routinely done statistical collection. As described previously, BJS will use the NCRP primarily to augment the National Prisoners Statistics to report annually on changes in the age, sex, race, and offense composition of the prison population at yearend, prison admissions during the year, prison releases during the year and expected time served upon admission. BJS will use the NCRP data in the production of three annual bulletins that are issued from the National Prisoners Statistics:

• Prisoners in 2011 (expected release October 2012)

• Prisoners in 2012 (expected release September 2013)

• Prisoners in 2013 (expected release August 2014)

By incorporating the NCRP into the Prisoners bulletins, BJS will achieve three objectives. It will provide national estimates of the characteristics of persons admitted into and released from prison in which the national estimates were produced from a methodology that address unit nonresponse in the NCRP (that is, states that do not report data) and will include estimates of imputation error. Previously published estimates from the NCRP were based on complete case data without estimates of imputation error. Using these estimates, BJS will be able to decompose changes in the composition of the prison population into sources of change, specifically admissions and time served. For example, BJS will be able to show how much of the change in the number of violent offenders in prison arose from changes in admissions or changes in time served by violent offenders. The same holds for other characteristics of the prison population. These statistics are consistent with the CNSTAT recommendation to improve the measurement of transitions between stages of the correctional system.

The NCRP age, sex, and race data are used to estimate the demographic distribution of state prisoners in the Correctional Population in the United States bulletins:

• Correctional Population in the United States, 2011 (expected release November 2012)

• Correctional Population in the United States, 2012 (expected release November 2013)

• Correctional Population in the United States, 2013 (expected release November 2014)

BJS also uses the NCRP in its work on Deaths in Custody (OMB # 1121-0249). Specifically, BJS uses the NCRP to create denominators in calculating mortality rates for the state prison population. Using the NCRP term records, BJS will be able to generate improved estimates of the average daily prison population by specific demographic characteristic for which mortality rates are calculated (e.g., age, race, sex, etc.). The reports that will utilize the NCRP include:

• Prison and Jail Deaths in Custody, 2000-2010 Statistical Tables (expected release October 2012)

• Prison and Jail Deaths in Custody, 2011 Statistical Tables (expected release October 2013)

• Prison and Jail Deaths in Custody, 2012 Statistical Tables (expected release October 2014)

Special topic reports

Four special topic reports that rely heavily on the NCRP data are planned:

• Aging prison population: This report will describe trends in the age distribution of the prison population over the past 30 years and focusing on the more current years, describe the differences between older prisoners (e.g., 55 plus) and those in younger age groups on offenses and time served. (Expected release, early 2013)

• Racial differences in all cause and cause-specific mortality among prisoners: This report will compare all cause and cause-specific, age-adjusted mortality rates in prison between racial groups and between prisoners and the general population. (Expected release, spring 2013)

• Temporal changes in imprisonment rates by race and sex: This report will look at changes in the distribution of state prisoners by race and sex, adjusting NCRP race data to account for states’ data systems’ inability to capture Hispanic ethnicity. (Expected release, spring 2014)

• Time served on life and other long sentences: Building upon the methodological work on time served, BJS will use the term records and hazard models to generate estimates of time to be served in prison for the types of offenses and offenders that enter on long sentences. Of interest to policy makers are questions related to actual time served on life sentences and whether “life means life.” (Expected release, spring 2015)

Online data dissemination tools

During the past several years, BJS has developed a series of online data dissemination tools that provide users with access to statistics through both static and dynamic query options. (See, for example, the N-VAT, an online tool for analyzing National Crime Victimization Survey (NCVS) data, available at:

http://bjs.ojp.usdoj.gov/index.cfm?ty=nvat.) To provide answers to routine requests for counts of prison populations and estimates of characteristics of the prison population, BJS plans to release a tool that will give users access to the combined NCRP and National Prisoners Statistics data. Users will be able to create customized reports based on the time period of interest to them, as well as demographic, offense, and sentence characteristics of the population. Longitudinal tables will let users track the prison population over time, and users will be able to download data in spreadsheet format. This online data tool is expected to be widely used by persons not wishing to perform these analyses themselves using the publicly available archived NCRP data. (Expected release mid-2013)

Data quality investigations that will result in reports

As described in the Needs and Uses section, BJS has underway a set of NCRP linking projects that are in progress and are expected to be completed during the next two years. These will result both in additional methodological and substantive papers. Given that they are relatively early in their development stage, expected dates for specific reports are yet to be determined, but the efforts and their likely substantive proceeds are described below:

• BJS will match NCRP prison release and parole discharge records with death records from the National Death Index (NDI) for the purpose of analyzing the mortality rate and life expectancy of prisoners and parolees released in 2009 and 2010. Two potential products will result: (1) a methodological document that reports on the results of the matching and yields a determination as to whether the linked data are reliable for estimation purposes; and (2) given that the linked data pass the test, a report on post-prison mortality that tracks time to death; crude and age-adjusted mortality rates for comparison the prison population; and cause-specific mortality rates, again for comparison with the prison population.

• Parole recidivism: While BJS preliminary analyses have proven the concept of linking NCRP records to estimate parole recidivism, as measured by return to prison, the linking needs to be completed for additional states. The requested enhancement variables for the NCRP-1C (parole discharge data) and the requested enhancement of the FBI number are required to complete this effort, as they allow BJS to include more states in the analysis of recidivism and assess out-of-state returns to prison. Once the methodological work to assess the quality of the entire linked dataset is complete, BJS plans to begin to issue on a routine basis estimates of parole recidivism.

• Labor markets and recidivism. BJS investigate enhancing the parole recidivism effort by linking data from the Bureau of Economic Analysis on county-level employment to the county of release of parolees and estimating the impact of demand for labor (e.g., employment rates) on recidivism. This effort would inform the reentry topic, as it will assess the effects of demand for labor on post-prison recidivism.

BJS Calendar for NCRP Documentation/Publications/Products

Type of publication/product	Title/topic of publication/product	Approximate date of release
Technical document	#1 – Observations on NCRP	December 2012 (to coincide with first archiving of the term record NCRP data)
Technical document	#2 – NCRP reporting	December 2012 (to coincide with first archiving of the term record NCRP data)
Technical document	#3 – A description of computing code used to identify correctional terms and histories	December 2012 (to coincide with first archiving of the term record NCRP data)
Technical document	#4 – Adjusting for unit nonresponse in NCRP	October 2013 (to coincide with archiving of term records with weights for adjustment to national estimates)
BJS Bulletin	Prisoners in 2011	October 2012
BJS Bulletin	Prisoners in 2012	September 2013
BJS Bulletin	Prisoners in 2013	August 2014
BJS Bulletin	Correctional Population in the United States, 2011	November 2012
BJS Bulletin	Correctional Population in the United States, 2012	November 2013
BJS Bulletin	Correctional Population in the United States, 2013	November 2014
BJS Special Report	Aging prison population	Early 2013
BJS Special Report	Racial differences in cause-specific age-adjusted mortality rates	Spring 2013
BJS Special Report	Temporal changes in imprisonment rates by race and sex	Spring 2014
BJS Special Report	Time served on life and other long sentences	Spring 2015
BJS Web Data Tool	NPS/NCRP data tool	Mid-2013

17. Expiration Date Approval

The OMB Control Number and the expiration date will be published on instructions provided to all respondents.

18. Exceptions to the Certification Statement

There are no exceptions to the Certification Statement. The collection is consistent with all the guidelines set forth in 5 CFR 1320.9.

1 National Research Council (2009). “Ensuring the Quality, Credibility, and Relevance of U.S. Justice Statistics. A Panel to Review the Programs of the Bureau of Justice Statistics”. Robert M. Groves and Daniel L. Cork, eds. Committee on National Statistics and Committee on Law and Justice, Division of Behavioral and Social Sciences and Education. Washington, DC: The National Academies Press.

2 Ingrid A. Binswanger, Marc F. Stern, Richard A. Deyo, Patrick J. Heagerty, Allen Cheadle, Joann G. Elmore, and Thomas D. Koepsell (2007). “Release from prison – a high risk of death for former inmates”, New England Journal of Medicine, Vol. 356, pp. 157-165.

3 Heather C. West, “Prison Inmates at Midyear 2009-Statistical Tables,” Bureau of Justice Statistics, Washington, DC, June 2010, NCJ 230113.

4 For a discussion of these issues related to time served and stable populations, see Evelyn J. Patterson and Samuel H. Preston (2007), “Estimating mean length of stay in prison: methods and applications”, Journal of Quantitative Criminology, Vol. 24, pp. 33-49.

5 The method—a piece-wise constant hazard model—takes into account all term records over a 10-year observation window and addresses both forms of censoring in the NCRP data. It takes into account truncation of admissions records and right censoring associated with admissions that have not yet been released.The principal authors of the paper—William Rhodes and Gerry Gaes—are, respectively, a senior methodologist at Abt Associates and the former director of the Office of Research and Evaluation at the Federal Bureau of Prisons.

6 See, for example Prisoners in 2010, tables 6 and 7 and appendix tables 7 and 8 for data on the gender, race, and age distributions of prisoners, and appendix tables 10 and 11 for the offense distributions. Earlier years of the Prisoners report also show these tables.

7 Evelyn J. Patterson (2010), “Measurement issues in corrections: quality and compatibility of state correctional data in the United States, 1984-2000”, Corrections Compendium, September.

8 Human Rights Watch (2008), “Targeting Blacks: Drug Law Enforcement and Race in the United States”, May.

9 Christopher Wildeman (2009), “Parental imprisonment, the prison boom, and the concentration of childhood disadvantage”, Demography, Vol. 46, Num. 2, pp. 265-280.

10 Alfred Blumstein and Allen J. Beck (1999), “Population Growth in U.S. Prisons, 1980-1996,” in Michael Tonry and Joan Petersilia, eds., Crime and Justice: Prisons, Chicago: University of Chicago Press, Vol. 26, pp. 17-61.

11 Steven Raphael (2009), “Explaining the Rise in U.S. Incarceration Rates,” Criminology and Public Policy, Vol. 8, No. 1, pp. 87-95.

12 William Spelman (2009), “Crime, Cash, and Limited Options: Explaining the Prison Boom,” Criminology and Public Policy, Vol. 8, No. 1, pp. 29-77.

13 John F. Pfaff (2011), “The myths and realities of correctional severity: evidence from the National Corrections Reporting Program on sentencing practices”, American Law and Economics Review, September, pp. 491-531.

14 Human Rights Watch (2012), “Old Behind Bars: The Aging Prison Population in the United States”, January.

15 Linh Vuong, Christopher Hartney, Barry Krisberg, and Susan Marchionna (2010), “The Extravagance of Imprisonment Revisited”, National Council on Crime and Delinquency, January.

16 Sara Markowitz, Erik Nesson, Eileen Poe-Yamagata, Curtis Florence, Partha Deb, Tracy Andrews, and Sarah Beth L. Barnett (2012), “Estimating the relationship between alcohol policies and criminal violence and victimization”, National Bureau of Economic Research Working Paper Series, Cambridge, MA.

17 Jeff Manza, Christopher Uggen, and Marcus Britton (2002), “The truly disenfranchised: Felon voting rights and American politics”, American Sociological Review, Vol. 67, pp.777-803.

18 Randa Embry and Phillip M. Lyons, Jr (2012), “Sex-based sentencing: sentencing discrepancies between male and female sex offenders.” Feminist Criminology, Vol. 72, Iss. 2, pp 146-162.

19 Anne C. Spaulding, Ryan M. Seals, Matthew J. Page, Amanda K. Brzozowski, William Rhodes, and Theodore M. Hammett (2009), “HIV/AIDS among inmates of and releasees from US correctional facilities, 2006: Declining share of epidemic but persistent public health opportunity, PLoS One, Vol. 4, Iss. 11, e7558.

20 Hope Corman, Dhaval M. Dave, Nancy E. Reichman, and Dhiman Das (2010), “Effects of welfare reform on illicit drug use of adult women”. National Bureau of Economic Research working paper 16072.

21 Alfred Blumstein and Allen J. Beck, (2005), “Reentry as a Transient State between Liberty and Recommitment,” in Jeremy Travis and Christy Visher, eds., Prisoner Reentry and Crime in America, New York, Cambridge University Press, pp. 50-79.

22 Michael Tonry and Matthew Melewski, (2008), “The Malign Effects of Drug and Crime Control Policies on Black Americans,” in Michael Tonry, ed., Crime and Justice: A Review of Research, Chicago: University of Chicago Press, Vol. 37, pp.1-44.

23 Shawn Bushway and Peter Reuter, (2008), “Economists’ Contributions to the Study of Crime and the Criminal Justice System, in in Michael Tonry, ed., Crime and Justice: A Review of Research, Chicago: University of Chicago Press, Vol. 37, pp. 389-451.

24 Two recent papers addressed this issue using NCRP data: Evelyn J. Patterson and Samuel H. Preston (2008), “Estimating Mean Length of Stay in Prison: Methods and Applications,” Journal of Quantitative Criminology, Vol. 24, pp. 33-49; and John F. Pfaff’s 2011 paper, cited in footnote 8.

25 William J. Sabol and James P. Lynch, (2003), “Consequences of Incarceration on Family Formation and Unemployment in Urban Areas,” in Darnell Hawkins and Samuel L. Myers, Jr., (eds.), Crime Control and Social Justice: The Delicate Balance, Greenwood Press, 2003

18