Download:
pdf |
pdfTable of Contents: Supplementary Materials
Appendix A: The Omnibus Crime Control and Safe Street Act of 1968
Appendix B: Whitepaper on the Use of NCRP as a Research Platform
Appendix C: Results of state survey on their ability to provide prisoners’ 9‐digit
social security number to NCRP
Appendix D: Phone script to introduce reporting year 2015 data collection for states
already submitting data to NCRP
Appendix E: Phone script to introduce reporting year 2015 data collection for states
that have not recently submitted data to NCRP
Appendix F: Introductory letter from BJS to data respondents for collection of 2015
NCRP data
Appendix G: Introductory letter from data collection agent to data respondents for
collection of 2015 NCRP data
Appendix H: Instructions for NCRP data submission, reporting year 2015
Appendix I: NCRP frequently asked questions fact sheet
Appendix J: Examples of follow‐up emails to 5 states seeking clarification on
NCRP data submitted
Appendix K: Whitepaper on estimating the proportion of the prison population
that will serve long sentences using NCRP data
Page 1
Page 6
Page 38
Page 41
Page 44
Page 48
Page 51
Page 53
Page 125
Page 129
Page 145
Appendix A
The Omnibus Crime Control and Safe Street Act of 1968
DERIVATION
Title I
THE OMNIBUS CRIME CONTROL AND SAFE STREETS ACT OF 1968
(Public Law 90-351)
42 U.S.C. § 3711, et seq.
AN ACT to assist State and local governments in reducing the incidence of crime, to increase the effectiveness,
fairness, and coordination of law enforcement and criminal justice systems at all levels of government, and for other
purposes.
As Amended By
THE OMNIBUS CRIME CONTROL ACT OF 1970
(Public Law 91-644)
THE CRIME CONTROL ACT OF 1973
(Public Law 93-83)
THE JUVENILE JUSTICE AND DELINQUENCY PREVENTION ACT OF 1974
(Public Law 93-415)
THE PUBLIC SAFETY OFFICERS’ BENEFITS ACT OF 1976
(Public Law 94-430)
THE CRIME CONTROL ACT OF 1976
(Public Law 94-503)
THE JUSTICE SYSTEM IMPROVEMENT ACT OF 1979
(Public Law 96-157)
THE JUSTICE ASSISTANCE ACT OF 1984
(Public Law 98-473)
STATE AND LOCAL LAW ENFORCEMENT ASSISTANCE ACT OF 1986
(Public Law 99-570-Subtitle K)
THE ANTI-DRUG ABUSE ACT OF 1988
TITLE VI, SUBTITLE C - STATE AND LOCAL NARCOTICS CONTROL
AND JUSTICE ASSISTANCE IMPROVEMENTS
(Public Law 100-690)
THE CRIME CONTROL ACT OF 1990
(Public Law 101-647)
BRADY HANDGUN VIOLENCE PROTECTION ACT
(Public Law 103-159)
VIOLENT CRIME CONTROL AND LAW ENFORCEMENT ACT OF 1994
(Public Law 103-322)
NATIONAL CHILD PROTECTION ACT OF 1993, AS AMENDED
(Public Law 103-209)
and
CRIME IDENTIFICATION TECHNOLOGY ACT OF 1998
(Public Law 105-251)
BUREAU OF JUSTICE STATISTICS
CHAPTER 46 - SUBCHAPTER III
[TITLE I - PART C]
42 USC § 3731
[Sec. 301.] Statement of purpose
It is the purpose of this subchapter [part] to provide for and encourage the collection and analysis of
statistical information concerning crime, juvenile delinquency, and the operation of the criminal justice
system and related aspects of the civil justice system and to support the development of information and
statistical systems at the Federal, State, and local levels to improve the efforts of these levels of government
to measure and understand the levels of crime, juvenile delinquency, and the operation of the criminal
justice system and related aspects of the civil justice system. The Bureau shall utilize to the maximum
extent feasible State governmental organizations and facilities responsible for the collection and analysis of
criminal justice data and statistics. In carrying out the provisions of this subchapter [part], the Bureau shall
give primary emphasis to the problems of State and local justice systems.
42 USC § 3732
[Sec. 302.] Bureau of Justice Statistics
(a) Establishment. There is established within the Department of Justice, under the general authority of the
Attorney General, a Bureau of Justice Statistics (hereinafter referred to in this subchapter [part] as
“Bureau”).
(b) Appointment of Director; experience; authority; restrictions. The Bureau shall be headed by a Director
appointed by the President, by and with the advice and consent of the Senate. The Director shall have had
experience in statistical programs. The Director shall have final authority for all grants, cooperative
agreements, and contracts awarded by the Bureau. The Director shall report to the Attorney General
through the Assistant Attorney General. The Director shall not engage in any other employment than that
of serving as Director; nor shall the Director hold any office in, or act in any capacity for, any organization,
agency, or institution with which the Bureau makes any contract or other arrangement under this Act.
(c) Duties and functions of Bureau. The Bureau is authorized to–
(1) make grants to, or enter into cooperative agreements or contracts with public agencies,
institutions of higher education, private organizations, or private individuals for purposes related
to this subchapter [part]; grants shall be made subject to continuing compliance with standards for
gathering justice statistics set forth in rules and regulations promulgated by the Director;
(2) collect and analyze information concerning criminal victimization, including crimes against the
elderly, and civil disputes;
(3) collect and analyze data that will serve as a continuous and comparable national social
indication of the prevalence, incidence, rates, extent, distribution, and attributes of crime, juvenile
delinquency, civil disputes, and other statistical factors related to crime, civil disputes, and
juvenile delinquency, in support of national, State, and local justice policy and decision making;
(4) collect and analyze statistical information, concerning the operations of the criminal justice
system at the Federal, State, and local levels;
(5) collect and analyze statistical information concerning the prevalence, incidence, rates, extent,
distribution, and attributes of crime, and juvenile delinquency, at the Federal, State, and local
levels;
(6) analyze the correlates of crime, civil disputes and juvenile delinquency, by the use of statistical
information, about criminal and civil justice systems at the Federal, State, and local levels, and
about the extent, distribution and attributes of crime, and juvenile delinquency, in the Nation and
at the Federal, State, and local levels;
(7) compile, collate, analyze, publish, and disseminate uniform national statistics concerning all
aspects of criminal justice and related aspects of civil justice, crime, including crimes against the
elderly, juvenile delinquency, criminal offenders, juvenile delinquents, and civil disputes in the
various States;
(8) recommend national standards for justice statistics and for insuring the reliability and validity
of justice statistics supplied pursuant to this chapter [title];
(9) maintain liaison with the judicial branches of the Federal and State Governments in matters
relating to justice statistics, and cooperate with the judicial branch in assuring as much uniformity
as feasible in statistical systems of the executive and judicial branches;
(10) provide information to the President, the Congress, the judiciary, State and local
governments, and the general public on justice statistics;
(11) establish or assist in the establishment of a system to provide State and local governments
with access to Federal informational resources useful in the planning, implementation, and
evaluation of programs under this Act;
(12) conduct or support research relating to methods of gathering or analyzing justice statistics;
(13) provide for the development of justice information systems programs and assistance to the
States and units of local government relating to collection, analysis, or dissemination of justice
statistics;
(14) develop and maintain a data processing capability to support the collection, aggregation,
analysis and dissemination of information on the incidence of crime and the operation of the
criminal justice system;
(15) collect, analyze and disseminate comprehensive Federal justice transaction statistics
(including statistics on issues of Federal justice interest such as public fraud and high technology
crime) and to provide technical assistance to and work jointly with other Federal agencies to
improve the availability and quality of Federal justice data;
(16) provide for the collection, compilation, analysis, publication and dissemination of
information and statistics about the prevalence, incidence, rates, extent, distribution and attributes
of drug offenses, drug related offenses and drug dependent offenders and further provide for the
establishment of a national clearinghouse to maintain and update a comprehensive and timely data
base on all criminal justice aspects of the drug crisis and to disseminate such information;
(17) provide for the collection, analysis, dissemination and publication of statistics on the
condition and progress of drug control activities at the Federal, State and local levels with
particular attention to programs and intervention efforts demonstrated to be of value in the overall
national anti- drug strategy and to provide for the establishment of a national clearinghouse for the
gathering of data generated by Federal, State, and local criminal justice agencies on their drug
enforcement activities;
(18) provide for the development and enhancement of State and local criminal justice information
systems, and the standardization of data reporting relating to the collection, analysis or
dissemination of data and statistics about drug offenses, drug related offenses, or drug dependent
offenders;
(19) provide for research and improvements in the accuracy, completeness, and inclusiveness of
criminal history record information, information systems, arrest warrant, and stolen vehicle record
information and information systems and support research concerning the accuracy, completeness,
and inclusiveness of other criminal justice record information;
(20) maintain liaison with State and local governments and governments of other nations
concerning justice statistics;
(21) cooperate in and participate with national and international organizations in the development
of uniform justice statistics;
(22) ensure conformance with security and privacy requirement of section 3789g of this title and
identify, analyze, and participate in the development and implementation of privacy, security and
information policies which impact on Federal and State criminal justice operations and related
statistical activities; and
(23) exercise the powers and functions set out in subchapter VIII [part H] of this chapter [title].
(d) Justice statistical collection, analysis, and dissemination. To insure that all justice statistical collection,
analysis, and dissemination is carried out in a coordinated manner, the Director is authorized to–
(1) utilize, with their consent, the services, equipment, records, personnel, information, and
facilities of other Federal, State, local, and private agencies and instrumentalities with or without
reimbursement therefore, and to enter into agreements with such agencies and instrumentalities for
purposes of data collection and analysis;
(2) confer and cooperate with State, municipal, and other local agencies;
(3) request such information, data, and reports from any Federal agency as may be required to
carry out the purposes of this chapter [title];
(4) seek the cooperation of the judicial branch of the Federal Government in gathering data from
criminal justice records; and
(5) encourage replication, coordination and sharing among justice agencies regarding information
systems, information policy, and data.
(e) Furnishing of information, data, or reports by Federal agencies. Federal agencies requested to furnish
information, data, or reports pursuant to subsection (d)(3) of this section shall provide such information to
the Bureau as is required to carry out the purposes of this section.
(f) Consultation with representatives of State and local government and judiciary. In recommending
standards for gathering justice statistics under this section, the Director shall consult with representatives of
State and local government, including, where appropriate, representatives of the judiciary.
42 USC § 3733
[Sec. 303.] Authority for 100 per centum grants
A grant authorized under this subchapter [part] may be up to 100 per centum of the total cost of each
project for which such grant is made. The Bureau shall require, whenever feasible as a condition of
approval of a grant under this subchapter [part], that the recipient contribute money, facilities, or services to
carry out the purposes for which the grant is sought.
42 USC § 3735
[Sec. 304.] Use of data
Data collected by the Bureau shall be used only for statistical or research purposes, and shall be gathered in
a manner that precludes their use for law enforcement or any purpose relating to a particular individual
other than statistical or research purposes.
Appendix B
Whitepaper on the Use of NCRP as a Research Platform
The NCRP Data as
a Research
Platform:
Evaluation Design
Considerations
Draft
May 28, 2015
Prepared by:
William Rhodes
Gerald Gaes
Ryan Kling
Jeremy Luallen
Tom Rich
Abt Associates
55 Wheeler Street
Cambridge, MA 02138
This work was supported by Grant No. 2010-BJ-CX-K067 awarded by the Bureau of Justice
Statistics, Office of Justice Programs, U.S. Department of Justice. Points of view in this document are
those of the authors and do not necessarily represent the official position of the U.S. Department of
Justice. This copy is not to be disseminated or cited.
Contents
1.
The NCRP as Panel Data ........................................................................................................ 2
2.
Defining Variables and Statistical Methodology .................................................................. 3
2.1 Terminology and Data Transformations .......................................................................... 3
2.1.1
Offense Seriousness............................................................................................ 4
2.1.2
Admissions ......................................................................................................... 5
2.1.3
Stocks ................................................................................................................. 6
2.1.4
Data Problems and Adjustments......................................................................... 6
2.1.5
Scaling for Visualization .................................................................................... 7
2.2 Regression Specifications ................................................................................................ 7
3.
Descriptive Statistics ............................................................................................................... 8
4.
Evaluation .............................................................................................................................. 16
4.1 Interrupted Time-Series ................................................................................................. 17
4.2 Difference-in-Differences .............................................................................................. 18
4.2.1
An Alternative Approach.................................................................................. 19
4.2.2
Multi-State Considerations ............................................................................... 20
4.3 Difference-in-Difference-in-Differences ....................................................................... 21
4.4 Synthetic Control Methodology .................................................................................... 22
5.
Conclusions ............................................................................................................................ 27
References ........................................................................................................................................... 28
Abt Associates
The NCRP Data as a Research Platform ▌pg. i
A hypothetical evaluation question posits that a state introduced a reform intended to reduce
incarceration for a targeted group of offenders. This white paper discusses how the Bureau of Justice
Statistics’ National Corrections Reporting Program (NCRP) data might be used to investigate what
that reform accomplished. Ultimately (but not in this paper) we seek to evaluate the Justice
Reinvestment Initiative (JRI), and given that ultimate goal, this paper is a design report.
From the modern framework of potential outcomes (Imbens & Rubin, 2015), evaluation always poses
a missing data problem. Once a state introduces a reform, an evaluator can observe what happened
following that introduction, but the evaluator cannot tell what would have happened had the state not
introduced that reform. The counterfactual is missing data.
The solution to the missing value problem is to compare the outcome following implementation of the
intervention with a selected counterfactual that presumably approximates what would have happened
absent the intervention. With qualifications (Berk, 2005), evaluators usually feel confident about
counterfactuals that are based on random assignment (Orr, 1999), but random assignment is
impractical for large-scale prison reforms. The alternative to random assignment is quasi-experiments
that exploit naturally occurring variation in what is sometimes called observational data. Quasiexperimental designs are tricky because they raise validity and reliability challenges.
This paper is a discussion of selected quasi-experimental approaches that should be useful for dealing
with the above evaluation question: pretest-posttest designs, difference-in-difference designs,
difference-in-difference-in-differences designs, and synthetic control methods. This is not an
exhaustive list of evaluation strategies, but we intend to emphasize the analysis of panel data (defined
below) derived from the NCRP. After examining these different evaluation approaches, we conclude
that each of the approaches has merit and that a thoughtful evaluation would exploit the advantages of
each.
As the argument advances, some definitions will be helpful.
Effect
A treatment effect or just effect is what the state actually accomplished because of the
reform. It might be defined as the reduction in the number or percentage of the
targeted population appearing in state prison relative to the size or percentage of that
targeted population that would have appeared absent the intervention.
Estimate
The above is definitional. The evaluator’s problem is to estimate the size of that
effect by identifying an appropriate counterfactual using procedures described in
many books concerned with evaluation (Lee, 2005; Cameron & Trivedi, 2005;
Angrist & Pischke, 2009; Rosenbaum, 2009; Morgan & Winship, 2015).
Validity
If the counterfactual does not provide a good comparison, we say that the evaluation
design poses a validity challenge, meaning that even in a very large sample, the
estimated effect would not approximate the real effect (Manski, 2007).
Reliability
Even if the counterfactual is valid, the amount of information provided by the data
may be so meagre that the estimated treatment effect is measured with great
imprecision. When the sampling variance for the estimated treatment effect is large,
we say that the evaluation design has little power or inadequate reliability.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 1
The question facing us is: How we can use the NCRP to estimate effects that are both valid and
reliable? Comparing the pre-implementation period with the post-implementation period within the
same state raises validity concerns because changes might have happened without the intervention.
Contrasting states that did and did not implement the intervention raises other validity concerns
because differences might have occurred for reasons other than the intervention. Furthermore,
reliability is challenging when performing state comparisons, because given a maximum of 50 states,
sample sizes are small.
Without pretense of being either comprehensive or final, this white paper walks through evaluation
design considerations specific to the NCRP. We illustrate use of those designs using NCRP data from
two states: Arizona and California. However, this paper does not provide an evaluation of policy
interventions in either state; we merely use these two to demonstrate how an evaluation might be
conducted. Arizona is a convenient choice because, to our knowledge, there has been no major policy
intervention within the state. Using Arizona data, we would expect that a demonstration evaluation
would find no effect of an imagined policy intervention. California is a convenient choice because its
prison Realignment initiative toward the end of the data assembly period had a widely acknowledged
effect on state prisons. Using California data, we would expect that a demonstration evaluation would
identify the effect from that known intervention.
This paper has five principal parts. The first discusses how the NCRP can be arranged into panel data;
this arrangement is especially useful for both description and evaluation. The second part introduces
some terms, describes some data transformations, and discusses statistical methodology exclusive of
evaluation methodology. The third part describes patterns in prison admissions and prison
populations in Arizona and California. This description is background for the discussion of evaluation
methodology, the focus of this paper, which appears in part four. Part five offers some concluding
remarks.
1.
The NCRP as Panel Data
Sponsored by the Bureau of Justice Statistics, the NCRP was redesigned beginning in 2010 to
assemble prison term records and post-confinement community supervision term records provided by
state authorities (Luallen, Rhodes, Gaes, Kling, & Rich, 2014).1 A prison term record begins when an
offender enters prison and ends when he or she leaves. The same offender may have multiple terms.
The records are updated yearly for each currently participating state and have been collected
retrospectively for some states that had not previously reported. Defined similarly, the assembly of
post-confinement community supervision (PCCS) records is a recent expansion of the NCRP; postconfinement records are not considered further in this paper although we could apply analogous
evaluation tools to PCCS.
1
The redesign was intended to increase state participation, improve data quality, and increase the data’s
utility for research. Previous users of the NCRP might note that the prison-term-based record arrangement
replaced the earlier reliance on unlinked admission and release records (A and B records) and stocks as of
December 31 of the reporting year (D records). The current NCRP allows stocks to be known for any date
within the observation window.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 2
The NCRP is designed to capture all prison terms that were active sometime during a window period
beginning in 2000 and ending (currently) in 2014. However, reporting patterns and data quality vary
by state. For many states, reporting is complete starting in 2000 and their data are deemed to be
sufficiently reliable so that the NCRP team could assemble prison term records for all reporting years.
Other states either report insufficient data (e.g., stock records but not admission and release records2)
or the reported data are deemed unreliable for one or more years starting in 2000. For these states, the
NCRP team either did not assemble term records at all or assembled term records beginning at some
year after 2000. Prison terms have also been assembled for Federal prisons (as part of BJS’s Federal
Justice Statistics Program), but those Federal records are not yet part of the NCRP. Because of
jurisdictional differences, it seems doubtful that Federal records would be useful counterfactuals for
evaluating state interventions.
When assembling descriptive statistics, and when explaining patterns in prison usage, assembling the
NCRP term file into panel data is helpful. In this paper, panel data comprise a cross-section of timeseries aggregates.3 Cross-sections are defined as states or frequently as offense combinations within a
single state. Time-series are months although other time-series units might be useful. Aggregates are
sums of units (such as admissions and prison stocks) or averages (such as average time-served). As an
illustration, picture measures of the number of admissions (the aggregate) for violent crimes, property
crimes and drug law violations within Arizona (the cross-sections) for every month between 2003 and
2012 (the time-series).
The analysis associated with this memo begins by using NCRP data from 2003 through 2012, a
period during which 26 states have prison term records. The analysis eventually reduces this
observation window because it turns out that most of the interesting trends happen after 2003, and by
starting the observation window later, we can include additional states in the analysis.
2.
Defining Variables and Statistical Methodology
This paper discusses evaluation methodology but preliminary to that discussion we define terms
whose meaning might otherwise be ambiguous. We also discuss the regression specification that
enters into the evaluation methodology. We do not discuss evaluation design per se in this section.
2.1
Terminology and Data Transformations
Three terms appear repeatedly in the rest of this paper.
2
It is possible to construct prison term records based on stocks alone provided a term is defined as lasting at
least one year. Additionally, when assembling descriptive statistics, we can impute missing terms from
stocks under an assumption that admissions appear steadily at the prison during a year. Because the limited
number of states restricts power, it might be useful to include additional states in the analysis despite this
limitation. However, we do not follow that route for this demonstration.
3
Panel data might be expressed as individual units (terms in our application), in which case the individual
units are the cross-section. For some purposes, analyzing the NCRP data at the individual level may be
insightful, but this paper is concerned with analyzing aggregate units so it adopts a narrow definition for
panel data.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 3
Offense seriousness: Correctional interventions frequently are targeted on a specific type of
offense or offender. For this paper, we presume the intervention targets offense types defined
by seriousness, and below we explain how we determined seriousness.
Admissions: Some interventions are best characterized as altering the rate at which offenders
enter prison.
Stocks: Other interventions are characterized as altering the prevalence of offenders in prison.
Admissions and stocks are examined on a per capita basis, which requires some data manipulation,
discussed below.
2.1.1
Offense Seriousness
Because correctional interventions often target offenses by seriousness, and because relative
seriousness is not obvious from an offense name, this paper creates offense seriousness categories.
Note that a useful definition of an offense category would depend on the intervention, so the
seriousness categories used here are purely illustrative. For example, an intervention targeted on druglaw violators would dictate a different way of defining offense categories.
Each prison term in the NCRP dataset is associated with a BJS offense code (assigned to the variable
BJS_Offense_1 in the NCRP). Using data from all states reporting to the NCRP since 2000, we
computed the mean time-served by individuals released from prison by offense code. (When
computing time-served, we excluded the records for offenders who served fewer than 90 days
because this exclusion allows us to adjust (imperfectly) for time-served following a revocation for a
technical violation.) Using average time-served, we placed every offender into a unique quintile
ranging from least to most serious offenses, i.e. from least to most time-served, on average. The
quantiles define five ordered seriousness categories.
Based on prison admissions, table 1 shows the distribution of seriousness categories cross-tabulated
with traditional generic offense groupings—violent, property, drug, other, and missing. Given the
remarkable dispersion of seriousness across offense types, we question how informative generic
offense types are for classifying data, but that is a topic for another time. We will use these offense
seriousness categories in this paper.
Table 1: Tabulation of Generic Offense Categories and Ascribed Seriousness
Categories
off_type
Violent
Property
Drug
Other
Missing
1
seriousness
2
3
4
5
7 93,184 4,975 365,922 618,879
270,240 556,971 9,582 446,504 34,112
375,161 257,429 500,579 63,497
1
270,091 129,766 249,963
299 18,062
244
48,556
Note that it is possible to include (or exclude) offense types and still classify by seriousness. For
example, just select violent offenses and compute seriousness categories within that grouping. We
Abt Associates
The NCRP Data as a Research Platform ▌pg. 4
suspect that this approach may place unwarranted weight on states having similar reporting
conventions, but that too should be a topic for research.
Classifying offenses by seriousness using time-served as an objective measure has some appeal for
understanding prison populations and comparing populations across states. Classification is especially
useful for evaluation because reforms often target a specific seriousness category (especially the least
serious crimes) for an intervention, suggesting that a counterfactual comes from comparing the
targeted population with the next less serious crimes (which should not be affected by the
intervention). This need for counterfactuals highlights the need for careful consideration of
seriousness categories. Three considerations seem important:
1. Many interventions identify the targeted category using a combination of offense type and
offender criminal history. The NCRP does not yet include any measure of criminal history
although it is possible to develop a proxy measure suitable for many analyses.4 Because our
concern is with demonstration, we have not attempted to apply this proxy in our analysis.
2. Useful evaluation requires careful thought about offense classification. For example, if the
state targeted offenders convicted of drunk driving, the counterfactual might be other crimes
that result in sentences roughly equivalent to the sentences for drunk driving. We employ the
seriousness categories for demonstration, not because they are necessarily the best way to
create counterfactuals for all evaluation questions, but because we are interested in
demonstration.
3. Both random assignment and quasi-experiments require the evaluator to justify the stable unit
treatment evaluation assumption (SUTVA). In the present context, SUTVA means that the
effect of the intervention does not spill over into the counterfactual comparison. For purposes
of discussion, we will maintain SUTVA, but a proper evaluation would carefully select the
comparison subjects to make SUTVA most plausible.5
2.1.2
Admissions
When assembling data, we discarded admissions when the term lasted for 90 days or fewer. This
choice is arguable but it eliminates short periods for revocations. The choice is also problematic in
that we cannot tell time-served for those who enter within 90 days of the final observed date so, for a
few states, there is a slight bias upward for admissions during the last 90 days of the observation
window. (That is, when no other information is available, we assume all terms with unobserved
4
The NCRP data begin for most states in 2000, so if the analysis begins in 2003, it is possible to distinguish
offenders who were released from prison during a three-year window before their current admission from
offenders who lacked a previous criminal history so measured. This is a crude but presumably effective
way to distinguish offenders based on criminal history. This paper does not demonstrate this application.
5
SUTVA is most credible when interventions are rule driven, which we expect to be the case with most
prison reforms. Morgan and Winship (2015) provide a helpful discussion of SUTVA and how to deal with
violations. For example, suppose an intervention targeted drunk drivers but some offenders convicted of
public intoxication (rather than drunk driving) are incidentally considered comparable and are released. The
evaluator might drop public intoxication from the comparison group and contrast drunk driving with other
offenses of comparable seriousness. Thoughtful consideration can mitigate or eliminate the SUTVA
problem.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 5
releases last 91 days or longer.) This bias will not be serious for this paper because most of the states
have reported 2013 data, and given 2013 data, we know when time-served lasted more than 90 days
for terms commencing in 2012.
2.1.3
Stocks
Our definition of stocks is just releases minus admissions for a given month. This is really the change
in stocks, but given the beginning stock in 2000, it is easy to compute cumulative stocks from
changes in stocks.6 For econometric analysis, dealing with changes in stocks (essentially a first
difference) has more desirable statistical properties than dealing with cumulative stocks.
2.1.4
Data Problems and Adjustments
The NCRP data have been matched with other data sources (Census data, FBI data, etc.) that provide
general population (age, arrests, etc.) statistics on a yearly basis. However, to capture interventions
that may have occurred during the year, we analyze prison statistics on a monthly basis, which causes
problems requiring adjustments.
Arrests
For example, consider prison admissions during January of 2005. If we hypothesize that prison
admissions are a function of arrests, we might regress admissions on arrests for 2005. The logical
problem is that while the admissions by construction occurred in January 2005, about 11 of every 12
arrests during 2005 occurred after January (and this ignores the delay from arrest to conviction to
incarceration), so the regression is misspecified.
Our approach is:
1. When analyzing year Y admissions in January, we use the weighted average of 11/12 year Y1 arrests and 1/12 year Y arrests.
2. When analyzing year Y admissions in February, we use the weighted average of 10/12 year
Y-1 arrests and 2/12 year Y arrests.
3. We make this adjustment progressively for other months.
This approach makes some strong assumptions about the lags between arrests and admissions, and a
refined analysis is required to develop an empirically justified distributed lag structure.7 We have not
done that for this discussion.
6
As noted earlier, the NCRP include all terms that were active sometime during the observation window.
This implies that an investigator can always construct the stock population on any date during that window
by a cumulative tabulation over time of admissions minus releases.
7
Our assumption is that arrests during the current month and arrests during the previous 11 months
contribute equally to admissions/stocks during the current month. An alternative would be to lag the effect
of arrests. For example, the previous 12 months (not including the current month) might account equally
for admissions/stocks. Or the previous months might have unequal weights so that arrests from 6 months in
the past have greater weight than 1 month and 12 months in the past. Possibly the lag structure should
extend longer than 12 months. Different lag structures are testable using the data to identify best fit but we
have not done that here.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 6
Population
For many purposes, it is instructive to examine admissions per capita or stocks per capita, but the
issue is “what should we use as population?” The current NCRP data report state population for the
year, and we adopt an adjustment similar to that used for arrests to distribute population over time.
However, this begs the question: Who is counted in the at-risk population? We adopted an expedient
approach of using the male and female population between 14 and 34; although 14 is too young for
prison admissions, we are constrained by Census-reported age categories.8
Scaling by population facilitates cross-state comparisons by accounting for population growth.
However scaling can distort raw trends. For example, prison population may increase on a raw basis
yet decrease on a per capita basis. Depending on the research question, scaling might be
inappropriate.
2.1.5
Scaling for Visualization
Another form of scaling is important for visualization. For some of the analysis, our approach is to
standardize change in stock by subtracting the mean change and dividing by the standard deviation.
Because it places statistics on a standard basis, this scaling facilitates drawing comparisons by crosssections. The application of this scaling will be obvious from the context because statistics will be
centered on zero and have a standard deviation of one.
2.2
Regression Specifications
Our analyses are always based on regressions even when the analysis is motivated by description. We
do not want to get too deeply into the details (which receive additional coverage in context) but:
1. To capture short-term patterns in trends, we use Fourier transformations that account for year
and half-year cycles. To capture long-term trends, we use polynomials. Specifically, Fourier
transformations use trigonometric functions (sine and cosine) to capture cycles that repeat
every year and half year.9 We do not know why these cycles occur, but we suspect they are
related to court cycles and delays between conviction and prison admissions. The cycles do
not much interest us, but accounting for them reduces residual variance so we can better see
what does interest us. When we use Fourier transformations, we first test for whether the year
and half-year effects are jointly statistically significant at p < 0.05. If not, we drop them from
the analysis; otherwise, we retain both the year and half-year effects.
2. Polynomials are useful for modeling long-term trends, the patterns that do interest us. Time is
always rescaled to run from 0 to 1 by dividing the months by 120, the total number of months
in the observation window. This rescaling helps with interpretation and does not alter the
8
The approach is expedient because older offenders are at risk of entering prison. An alternative approach
would be to weight the age groupings according to the age of offenders entering prison. We have not taken
that step in this paper.
9
Fourier transformations are sometimes uses to capture cyclical behavior because a Fourier transformation
can capture any repeated pattern with an arbitrary degree of precision. Our application requires four
terms—a sine and cosine function that repeats on a yearly basis and a sine and cosine function that repeats
on a half-year basis. Hence the regression shows four terms f1 through f4.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 7
regression results.10 When we use polynomials, we always start with a cubic. A polynomial
based on a cubic includes time, time-squared and time-cubed. When we use a polynomial, we
first test whether the cubed term is significant at p < 0.05. If it is not significant, we drop the
cubed term and test for the squared term, and if that is not significant, we then test for the
linear trend. If it is not significant, there is no trend.
Other variables are incorporated into the regression. Seeking to demonstrate techniques, we have not
attempted to be comprehensive. The arrest variables (and sometimes lagged releases) enter into some
of the regressions. Typically we perform a joint test for statistical significance, and if the variables are
not statistically significant, we drop them (at p < 0.05).
Dependent variables are scaled by dividing by population and sometimes additional units (such as
division by 100) to provide interpretable pictures. Regression parameters are difficult to interpret and
we suggest examining them qualitatively (for direction) but ignoring them quantitatively (for
magnitude). Because of collinearity, even qualitative interpretations can be uninformative, so the
reader might treat collinear variables as just “adjusting” for past arrests; collinearity will not affect the
joint explanatory power of even perfectly collinear variables.
3.
Descriptive Statistics
Descriptive analysis is a useful starting point. We show figures summarizing long-term trends in two
states that have reported to the NCRP since 2003. The purpose of presenting descriptive trends is
simply to illustrate considerable fluctuation in stocks (prison stock) and flows (admissions) over short
periods of time. These fluctuations complicate evaluation because, when short-term changes occur
naturally, interrupted time-series are unreliable for forming counterfactuals. Given this limited
purpose, we only show trends for Arizona and California, two states that are the focus when this
paper turns from description to illustrating approaches to evaluation.
Figure 1 shows Arizona admissions per 100,000 residents between 16 and 34, in total and broken
down by offense seriousness category. The figure has six panels corresponding to the five seriousness
classes and all classes rolled together. The dots are actual data. Table 2 shows regression results. If
the cycles were insignificant, then the curve would be smooth. Therefore, by just looking at the
figure, we can tell that the Fourier transformations are statistically significant except for seriousness
class 4. The cycles might exist for seriousness class 4, but power is insufficient to detect the pattern.
Regardless, unless the line is flat (perhaps with cyclical perturbations), we can tell whether the
polynomial is statistically significant. A sharp eye can even tell which degree of the polynomial is
statistically significant. There are strong seasonal and long-term trends in Arizona.
10
The regressions used here are invariant (except for scale effects) to linear transformations. Although a
polynomial may seem nonlinear, it is actually linear in its arguments, which is sufficient for the invariance
properties to hold.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 8
Predicted Admissions by Seriousness Class per Population
Arizona (monthly 2003-2012)
Seriousness class
3
24
48
72
month
96
120
0
24
48
72
month
96
120
24
48
72
month
96
120
Seriousness class
5
6 8 101214
6
8 10 12
Seriousness class
4
0
0
24
48
72
month
96
0
24
120
48
72
month
96
120
96
120
All classes
6 0 7 0 8 0 9 01 00
0
20 25 30 35
1 5 2 02 5 3 0 3 5
Seriousness class
2
6 8 1 01 21 4
Seriousness class
1
0
24
48
72
month
Predicted Monthly Admissions per 100000 population age 15-34
Based on a polynomial regression with Fourier transformations, lagged arrest and lagged releases.
Figure 1: Trends in Prison Admissions per Capita in Arizona
Because the figures are adequately descriptive, the regression parameters (table 2) are relatively
uninteresting. The polynomials are captured by the T, Tsq and Tq terms. The Fourier transformations
are captured by the f1 through f4 terms. If parameters appear in the table, then the polynomial/cycles
are statistically significant at p < 0.05.11 That is, the table indicates the degree of the polynomial used
to estimate the regression and whether the Fourier transformations entered the regression. The table
shows that past arrests are important for explaining admissions; the arrest variables would not appear
in the table if they were not jointly significant. Lagged releases are typically not statistically
significant. Except for seriousness class 4 admissions, the R2 gives an impression of substantial
change in admissions per capita over time. This is a context where R2 tells us little. If there are no
cyclical patterns and no trend, then the R2 would be near zero. An R2 of zero does not mean that we
have explained nothing; on the contrary, we have explained much—namely, there is no discernable
trend.
11
The table also shows which specific parameters are statistically significant, but the significance of
individual parameters should be of little interest. Joint tests are most interesting but not shown in the table.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 9
Table 2: Arizona Polynomial Regression
T
Tsq
viol
prop
drug
f1
f2
f3
f4
lagged_releases5
_cons
R2
N
admissions1
admissions2
admissions3
admissions4
admissions5
22.361**
-21.040**
2,028.386
123.841
-1,002.431**
-0.129
-0.207
-0.367**
-0.023
20.800**
-27.090**
12,336.912**
-1,393.047**
-1,058.146**
-0.605*
0.061
-0.873**
0.398
43.494**
-43.617**
11,082.520**
-1,340.379**
-1,651.783**
-0.403
0.063
-1.059**
0.267
5.807**
-5.653**
3,967.658**
-397.512**
-320.935*
11.830
0.71
120
12.824
0.70
120
25.060
0.63
120
1.406
0.23
120
13.074**
-11.562**
2,264.606*
166.634
-100.615
-0.218
0.101
-0.543**
-0.096
0.054**
-14.172
0.47
108
total
107.886**
-111.826**
32,503.767**
-2,931.792**
-4,629.464**
-1.373
0.263
-3.105**
0.507
50.566
0.58
120
* p<0.05; ** p<0.01
For present purposes, the story behind the trends in Arizona is simple. There are short-term
fluctuations and long-term reversals in trends. If we attempted to evaluate a policy intervention in
Arizona, these short-term fluctuations and long-term shifts would raise validity concerns. We return
to this point later.
Polynomials can give distorted impressions when admission practices suddenly shift. California
(Figure 2) illustrates this. California had been experiencing a decrease in prison population per capita
before it changed its admission practices (called Realignment) to make greater use of county jails.
The polynomial suggests a downward trend that really has abated by the last year of data, but the
polynomial does not show that subsequent abatement. (A higher degree polynomial might be helpful,
but probably a spline recognizing the known break in California admissions would be more helpful.)
Notice the high R2; these occur because of the precipitous drop in admissions, not because the
regressions really explain better in California than in Arizona.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 10
Predicted Admissions by Seriousness Class per Population
California (monthly 2003-2012)
24
48
72
month
96
120
8 10
6
4
0
24
48
72
month
96
120
24
48
72
month
96
120
Seriousness class
5
6 8 1 01 2 1 4
Seriousness class
4
0
0
24
48
72
month
96
0
24
120
48
72
month
96
120
96
120
All classes
2 0 4 0 6 0 8 0 1 00
0
0
5 1 0 1 5 2 02 5
Seriousness class
3
4 6 8 1 01 2
Seriousness class
2
10 20 30
Seriousness class
1
0
24
48
72
month
Predicted Monthly Admissions per 100000 population age 15-34
Based on a polynomial regression with Fourier transformations, lagged arrest and lagged releases.
Figure 2: Trends in Prison Admissions per Capita in California
Table 3: California Polynomial Regression
T
Tsq
Tq
viol
prop
drug
f1
f2
f3
f4
_cons
R2
N
admissions1
admissions2
admissions3
admissions4
admissions5
total
-58.610**
130.626**
-88.539**
59.711
-1,529.135*
1,832.610**
-0.997**
-0.153
-0.324
-0.045
5.349
0.95
120
-28.160**
66.244**
-39.915**
-374.048
289.447
755.396**
-0.465**
-0.087
-0.149
0.062
-8.119
0.89
120
-59.884**
141.910**
-90.954**
-469.230
-297.389
1,473.717**
-1.042**
0.005
-0.337
-0.070
-3.623
0.91
120
-19.684**
48.809**
-29.415**
604.787
-104.952
567.860**
-0.391**
-0.053
-0.115
-0.102
-10.672
0.88
120
-31.378**
81.263**
-52.312**
259.377
-13.903
571.489**
-0.526**
0.035
-0.287*
-0.100
-4.225
0.81
120
-197.717**
468.851**
-301.135**
80.597
-1,655.931
5,201.072**
-3.421**
-0.254
-1.211
-0.254
-21.290
0.92
120
* p<0.05; ** p<0.01
California offers a useful contrast to Arizona. In Arizona, the figure shows short-term fluctuations
and long-term reversals in trends; by assumption, made for purposes of this discussion, neither could
be attributed to a statewide intervention. If we had attempted to evaluate an intervention, these
naturally occurring changes would raise validity issues. In California, we know that the state
Abt Associates
The NCRP Data as a Research Platform ▌pg. 11
correctional system underwent a profound policy change, shifting offenders from state prisons to
county jails. Interesting, however, a cynical evaluator could point out that the post-intervention trends
appear to be an extension of preexisting trends. Descriptive statistics provide an inadequate platform
for evaluation.
Additional descriptive analysis comes from examining the monthly change in stocks beginning in the
first month (i.e., January 2003) and ending in December 2012. Monthly change—the first difference
of the cumulative change—is more useful for understanding trends because it more clearly relates
changes to covariates. That is, if we wanted to analyze changes in stocks, serial correlation would be
severe, so we would take first differences to reduce the serial correlation. That step is taken here.
Figure 3 shows actual data (the dots) and predictions (the lines) for Arizona. The change in the stock
of prisoners is the difference between admissions and releases in each month, so in theory this new
figure might tell us something different than did its admissions counterpart, but in fact the story does
not much change. As before, we see fluctuations in the change in the stock, cycles and long-term
shifts in trends. Basing an evaluation on an interrupted time-series would be tenuous.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 12
Predicted Stock Change by Seriousness Class per Population
Arizona (monthly stock change 2003-2012)
Seriousness class
2
Seriousness class
3
24
48
72
month
96
120
10
5
0
0
48
72
month
96
120
Seriousness class
5
2
0
-2
0
24
48
72
month
96
120
0
24
48
72
month
96
0
24
48
72
month
96
120
96
120
All classes
-2 0 2 4 6
4
Seriousness class
4
24
-1 0 0 1 0 2 0 3 0
0
-5
-4 -2 0 2 4
- 5 0 5 1 01 5
Seriousness class
1
120
0
24
48
72
month
Predicted Monthly stock change per 100000 population age 15-34
Based on a polynomial regression with Fourier transformation, lagged arrests and lagged releases.
Figure 3: Trends in per Capita Changes in Stocks in Arizona
Table 4: Regression Results for Trends in per Capita Changes in Stocks in Arizona
stock1
T
Tsq
lagged_releases1
f1
f2
f3
f4
Tq
viol
prop
drug
lagged_releases4
lagged_releases5
_cons
R2
N
14.193**
-9.446**
-0.037**
-0.119
-0.317
-0.474**
0.002
stock2
stock3
stock4
-1.579
-31.970
-6.990**
-3.136**
-0.558
0.072
-1.039**
0.961**
27.635*
-0.106
0.124
-1.161**
0.130
stock5
total
-18.081**
35.856**
-16.063**
-0.232
-0.006
-0.680**
-0.055
-25.672**
-1.033
0.243
-3.542**
1.072
6,570.743**
-1,421.503**
-547.111*
21,662.546**
-3,479.537**
-1,446.532**
0.047**
1.856
0.31
108
5.772**
0.60
120
16.311
0.45
120
-3.164*
0.16
108
0.089**
-3.442*
0.46
108
12.771
0.60
120
* p<0.05; ** p<0.01
Abt Associates
The NCRP Data as a Research Platform ▌pg. 13
Figure 4 is the counterpart to figure 2 for California. Although the story might have been different
from that told by admissions in California, in fact the story is quite similar. We see fluctuations and
cycles but no large interruptions in the trend except for the drastic drop in stocks following
California’s policy intervention. Late in the period, the change in stocks has hovered around zero,
much as it had during the years prior to the intervention. Even the cynical evaluator, identified earlier,
might find this abrupt change immediately after the intervention compelling; still, it would be helpful
to have a formal test. This concern brings us to the transition between descriptive statistics and
inferential statistics used for evaluation, the topic of the next section.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 14
Predicted Stock Change by Seriousness Class per Population
California (monthly stock change 2003-2012)
24
48
72
month
96
120
-1 0 -5
0
48
72
month
96
120
Seriousness class
5
0
24
48
72
month
96
120
0
24
48
72
month
96
0
24
48
72
month
96
120
96
120
All classes
-4 -2 0 2 4
- 3- 2- 1 0 1 2
Seriousness class
4
24
-3 0-2 0-1 0 0 1 0
0
0
-4 -2 0
5
Seriousness class
3
2
Seriousness class
2
- 15- 10 - 5 0 5
Seriousness class
1
120
0
24
48
72
month
Predicted Monthly stock change per 100000 population age 15-34
Based on a polynomial regression with Fourier transformation, lagged arrests and lagged releases.
Figure 4: Trends in per Capita Stocks in California
Table 5: Regression Results for Trends in Per Capita Stocks in California
stock1
T
Tsq
Tq
viol
prop
drug
f1
f2
f3
f4
lagged_releases5
_cons
R2
N
stock2
stock3
stock4
-71.510**
144.079**
-51.795**
-114.764
1,939.548**
2,439.854**
-1.474**
-0.084
-0.381
0.044
-25.467**
52.120**
-13.774*
1,265.110**
728.416**
927.589**
-0.573**
-0.048
-0.266*
0.064
-52.572**
110.051**
-38.823**
870.502
1,211.545*
1,972.270**
-1.315**
0.203
-0.280
0.113
-9.449**
18.662**
1,230.239**
328.098
582.595**
-0.444**
0.002
-0.142
0.009
-90.733**
0.69
120
-50.439**
0.51
120
-79.699**
0.67
120
-35.503**
0.49
120
stock5
-14.066**
26.834**
1,413.594*
295.366
919.159**
-0.560**
0.179
-0.384**
-0.087
0.004
-48.842**
0.45
108
total
-172.254**
352.982**
-108.210**
4,728.302
4,514.428*
6,672.004**
-4.367**
0.238
-1.419*
0.168
-297.535**
0.65
120
* p<0.05; ** p<0.01
Abt Associates
The NCRP Data as a Research Platform ▌pg. 15
4.
Evaluation
The rudiments of evaluation appear in the discussion above (that is, our eyes can detect patterns), but
formal designs are required to meet validity and reliability challenges. We discuss four evaluation
designs: interrupted time-series; difference-in-differences; difference-in-difference-in-differences;
and synthetic control methods. Throughout this discussion, the motivational illustration is that a state
decides to reduce its prison population for the least serious offenders. This policy shift occurs at a
defined point in time, although we might assume that the intervention takes time to reach full
implementation so the full effect is lagged.
Let:
S ijk
This is the stock of offenders from seriousness category i at time j in state k.
s ijk
This is the change in the stock from seriousness category i at time j in state k.
sijk S ijk S i ( j 1) k
Mj
This is the month, typically parameterized to run from 0 to 1 by dividing months by the
number of months in the observation window as described above. When drawing figures, to
assist the reader, we revert to using the months rather than transformed version of months.
These are all variables that we used above when presenting descriptive statistics.
The discussion of design in the remainder of this section is progressive. That is, the interrupted timeseries is the least useful and the synthetic estimation is arguably the most useful, but they actually
have much in common, so value comes from building more sophisticated approaches onto the less
sophisticated approaches. As the term is used here, an approach is more sophisticated if it raises fewer
validity concerns.
Although we derived the descriptive statistics from 2003–2012, based on the descriptive statistics we
doubt that such a long time-series is useful for evaluation because perturbations and reversals in trend
that occur early in the time-series are likely uninformative about interventions that occur later in the
time-series. Consequently, in the following demonstration, we will abbreviate the time-series. This
has the additional advantage of allowing us to expand the number of states under study. An evaluation
of the JRI would probably benefit from even a shorter observation window.
As a road map of the following subsections, for Arizona we imagine an intervention that happened
exactly two years before the end of the NCRP time-series. In fact, there was no intervention on that
date, so we would not expect to observe an effect. We then discuss using an interrupted time-series
(section 4.1), a difference-in-differences design (section 4.2) and a difference-in-difference-indifferences design (section 4.3) to “evaluate” this imagined intervention. The point is that the least
rigorous design can lead to spurious conclusions and the more sophisticated designs are more
believable. For California, a real intervention occurred toward the end of the time-series, purposefully
substituting confinement in county jails for confinement is state prisons. We use the synthetic case
control method to detect the consequences of that policy change (section 4.4).
Abt Associates
The NCRP Data as a Research Platform ▌pg. 16
4.1
Interrupted Time-Series
For Arizona we hypothesize a break in a trend on January 1, 2011 only for the least serious offenses,
which are assumed to be the target of the intervention. An approach to an interruptive time-series is to
assume that trends are linear or nearly linear immediately to the left and immediately to the right of
the break.12 The “treatment effect” is the shift in the regression lines at the intervention point. The
typical application selects a bandwidth (of time) that is clustered about the intervention point. Without
more discussion, we limit the analysis to two years before the intervention and two years after the
intervention. Given yearly cycles, bandwidths should always be specified as years. In practice we
would test alternative bandwidths, but this testing is not important for this demonstration.
We have standardized the stock by subtracting the mean change and dividing by the standard
deviation for the pre-intervention period. Without standardization, difference-in-differences and
difference-in-difference-in-differences comparisons are difficult to discern. With standardization,
statistics are centered near zero and have a standard deviation near one regardless of the original
scale.
Using the Arizona data, we fit a linear model in time to the left and a linear model in time to the right
of the intervention point. This model also includes cycles, and they are very important over this short
interval, but we will not show them because they dominate the picture. See figure 5. It shows the
predictions, based on the linear model after removing (partialing out) the cycles, for all five offense
seriousness categories (SC 1 through SC 5), but current attention is just on the first offense
seriousness category (SC 1).
12
Although the point is arguable, some evaluators treat an interrupted time-series as being a regression
discontinuity design (Imbens & Lemieux, 2007). From the RDD perspective, the estimated treatment effect
is most valid when it is estimated immediately about the break point using local linear regressions. The
RDD—and hence the interrupted time-series—has less appeal when the impact of an intervention
materializes over a lengthy period, one of the points made in this paper. Within a criminal justice context,
some of these issues are discussed in Rhodes and Jalbert (2013).
Abt Associates
The NCRP Data as a Research Platform ▌pg. 17
Changes in stocks: Linear spline about a hypothezied break
F itte d v a l u e s
- 1 .5 - 1 - .5 0
.5
1
Arizona: SUR regression with cycles partialed out
70
80
90
100
110
120
month
SC 1
SC 3
SC 5
SC 2
SC 4
Stock changes standardized using pre-intervention data
Figure 5: An Interrupted Time-Series for Changes in Per Capita Prison Stocks in
Arizona
Focusing our attention on seriousness class 1, the visual impression is that the stock increased at the
time of the imaginary intervention (i.e. after 24 months) and that the previously decreasing trend
reversed its course. Because there was no actual intervention, we expected to see a continuation of the
pre-24 month trend. In fact, the jump after 24 months is not statistically significant, but the reversal in
the trend is highly significant (p = 0.02); based on these results alone, we would falsely conclude that
our imaginary intervention changed the trend. In fact, looking across all five seriousness classes, the
jump is significant (p = 0.001) for one class and the reversal in trends is significant at p < 0.05 for two
seriousness classes and insignificant at p < 0.06 for a third. These results illustrate that resting
evaluation on an interrupted time-series is treacherous and raises validity concerns, causing us to
recommend against using an interrupted time-series to evaluate policy interventions intended to affect
populations in state prisons.13
4.2
Difference-in-Differences
A problem with the interrupted time-series is that the post-intervention period may differ from the
pre-intervention period for reasons that have nothing to do with the intervention. One way to
strengthen the inference about treatment effectiveness is to presume that, absent an effective
intervention, whatever changes occur during the post-intervention period would affect both
seriousness class 1 and seriousness class 2 crimes in approximately the same way. This implies that
we should compare the difference in trends for seriousness class 1 and seriousness class 2 crimes and
only reject the null of no treatment effect when the break/trend for seriousness class 1 crimes differs
13
An evaluator might choose to use a polynomial instead of local linear regressions, but this approach raises
validity issues. When the regression is nonlinear in the vicinity of the break point, distinguishing between
naturally occurring nonlinearity and nonlinearity induced by a true intervention is tenuous. We concede that
other evaluators may prefer a nonlinear regression nevertheless, and rather than argue the point, we just
emphasize that an interrupted time-series raises difficult problems of interpretation.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 18
from the break/trend for seriousness class 2 crimes. A similar logic might be employed to contrast
seriousness class 1 and seriousness class 3 crimes. This type of comparison is an application of a
difference-in-differences (DD) design. Note that this approach depends critically on being correct
about SUTVA so in a real application evaluators would be especially careful about choosing the cross
sections.14
There is a trick to deriving the standard error for the test statistic because the time-series are not
independent. We have used a linear seemingly unrelated regression (SUR) to estimate covariances.
Variances are unaffected because, for each seriousness class, the right-hand-side variables are the
same.
We compare the break in the time-series for seriousness class 1 with the break in the time-series for
seriousness class 2 and find no statistically significant difference. We compare the break for SC 1
with the break for SC 3 and again find no significant break. We compare the break for SC 1 with the
average of SC 2 and SC 3 and again find no statistically significant difference. Using these same
contrasts for the post-intervention trends, we find no statistically significant differences. The DD
design provides more satisfying results both because we fail to reject the null (which we know is
correct) and because the logic of a DD is more compelling than the logic of an interrupted time-series.
4.2.1
An Alternative Approach
Although the DD framework specified above is familiar, an alternative that uses essentially the same
identification strategy may be better. The alternative uses a ratio:
rijk
Sijk
S
i 1
ijk
The numerator is stock for the seriousness class that is targeted by the intervention. The denominator
is some combination of seriousness classes that are not targeted for the intervention. As before, the
denominator should probably be restricted to seriousness classes that are similar to the seriousness
class of interest.
We can substitute the ratio into the same regression framework used earlier for the interrupted timeseries. Because we have not taken a first difference, autocorrelation is a problem, and consequently
we have introduced a Prais-Winston transformation to adjust the regression for autocorrelation.
Figure 6 shows four ratios, over an abbreviated observation window, for Arizona. The highest curve
shows the ratio of class 1 seriousness offenders to the sum of class 1 and class 2 seriousness
offenders. The lowest curve shows the ratio of class 1 seriousness offenders to the sum of all
offenders. Visual inspection of the figure suggests no strong sharp breaks at 96 months. The evidence
is less compelling regarding trends, and in fact, the trends are statistically different during the
hypothetical intervention period (when no intervention in fact occurred) for two of the four contrasts.
However, there is no statistically significant change in the trends for SC 1/(SC 1 + SC 2) or for
14
Often interventions are rule driven, such as: release drug-law violators convicted of minor trafficking
offenses. A suitable comparison group would be offenders convicted of low-level property crimes or minor
assaults. The most desirable comparison group depends on the context so our choice of seriousness classes
is only for illustration.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 19
SC1/(SC1 + SC2 + SC 3); these are the contrast that seem most justified because SC 4 and SC 5
crimes are very different than SC 1 crimes. Even if we decide to place some emphasis on ratios that
are statistically significant, we note that the size of the effect is not substantively large, so effects
might be statistically significant but not substantively important. Given that the more proximate
seriousness classes are the most informative for SC 1, we put more faith in the comparisons for these
first two ratios.
Difference-in-Differences Based on Changes in Four Ratios
.0 5
L in e a r p r e d i c t io n
.1
.1 5
.2
.2 5
Arizona
70
80
90
100
110
120
month
SC 1/(1+2)
SC 1/(1+2+3+4)
SC 1/(1+2+3)
SC 1/(1+2+3+4+5)
Ratios are the ratio of stock of serious category 1 to sums of other categories
Figure 6: An Application of a Difference-in-Differences Estimator for Changes in
Prison Stocks in Arizona
The difference-in-differences approach does not eliminate validity challenges and a rigorous
evaluation might more carefully construct and examine the contrasts. The simple point made in the
above two figures is that a difference-in-difference design greatly reduces validity challenges that
arose in the interrupted time-series approach. A reader might think of the difference-in-difference
approach as starting with an interrupted time-series (which is an obvious element of the DD) and
improving the credibility of the inference.
4.2.2
Multi-State Considerations
Noteworthy, the DD estimates in the above two sections are state-specific. For some evaluations,
including an evaluation of the JRI, several states have implemented the interventions at about the
same time, and we might be interested in comparing effects across states or in combining effects to
get an average effect. Because the estimate from one state is independent of the estimate from another
state, we can combine estimates using an approach generically known as meta-analysis. There are
many sources explaining meta-analysis (Borenstein M. , Hedges, Higgins, & Rothstein, 2009) but the
simple analytics, suitable for present purposes, appear in accessible sources (Borenstein M. , Hedges,
Higgins, & Rothstein, 2010; Rhodes, 2012).
Basically if we derive treatment effects for N states, then we can average across those N states to
derive a composite treatment effect. We will not discuss the details, but a meta-analysis approach
leads to an average (or, as appropriate, a weighted average) with standard error that depend on
(among other things) whether the chosen estimator is a fixed-effect or a random-effect estimator.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 20
An important consideration when evaluating JRI is that JRI is not a unique program. Rather, the JRI
approach varies across the state implementers. This suggests that the meta-analysis might attempt to
explain the size of treatment effects using specific program components as explanatory variables.
Technically, this approach is a straightforward regression problem (Rhodes, 2012). Practically,
however, inferences are limited by (1) the small number of states that have adopted JRI and (2) the
diversity of JRI practices across those states. Rather than identifying an average treatment effect (a
fixed-effect model), it may be useful to identify the variance in treatment effects across settings (a
random-effect model) even it that variance cannot be explained due to insufficient data.
The difference-in-differences designed matched with meta-analysis appears to be a strong design for
evaluating correctional interventions that are targeted on a specific group of offenders. We
recommend combining DD and meta-analysis as the basic approach to dealing with state-level
correctional interventions. However, this paper discusses an additional approach—a difference-indifference-in-differences design—that may be suitable in some circumstances.
4.3
Difference-in-Difference-in-Differences
Although the DD framework appears to provide a useful basis for evaluation, we might strengthen
that inferential framework using a difference-in-difference-in-differences design, hereafter DDD. The
logic is that we compare the changes in slope in the state that implemented the intervention (Arizona,
here) with the changes in slope for states that did not implement the intervention.15 This is a DDD
design because the first difference is within state (using the ratio approach) and the second difference
is across state.
There are a number of ways to specify a model, but they all have a flaw: What other states should be
used in the comparison? This question used to receive little attention in econometrics. Recently it has
been receiving widespread attention (Wooldridge, 2007; Abadie, Diamond, & Hainmueller, 2010;
Abadie, Diamond, & Hainmueller, 2014; Imbens & Wooldridge, 2009).
The basic problem is that statistical testing assumes that the state or states used in the DDD
comparison are in fact appropriate comparisons, so that measurement error comes exclusively from
time-series fluctuations. In fact, if there are differences across states that are not taken into account by
matching states, then an additional level of uncertainty—that attributable to selecting the comparison
states—is incorrectly ignored by the analysis.
Rather than performing a mock evaluation with cross-state comparisons, we use descriptive statistics
from Georgia to illustrate the potential danger of a cross-state comparison. Compare figure 7
(Georgia) with figure 1 (Arizona). The presumption is that neither state had implemented major
interventions to alter prison admissions. The logic of a difference-in-difference (or difference-indifference-in-differences) methodology is that, for Georgia to be a useful counterfactual, both states
should show comparable trends absent any interventions, but clearly the comparison shows that
15
This is one way of testing the SUTVA. Returning to the earlier example, suppose the intervention affects
drunk driving but that offenders convicted of public intoxication might be treated similarly, perhaps
because they were actually charged with drunk driving but entered a plea to public intoxication. We would
expect to see a different trend for drunken driving in the state that implemented the intervention than in the
state that did not implement the intervention. If there was no spillover, we would not expect to see the same
contrast for public introxication.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 21
equivalency to be erroneous. Any comparison of trends in Arizona and Georgia would fail to capture
the uncertainly of selecting a state (Georgia) for purposes of comparison (with Arizona). Simply put,
Georgia is a poor counterfactual but how would an evaluator know that?
Predicted Admissions by Seriousness Class per Population
Georgia
Seriousness class
3
24
48
72
month
96
120
48
72
month
96
120
0
24
48
72
month
96
120
0
24
48
72
month
96
0
24
48
72
month
96
120
96
120
All classes
5
10 15 20
5
24
Seriousness class
5
10 15 20
Seriousness class
4
0
4 05 06 07 08 09 0
2 3 4 5 6
0
5 1 01 5 2 0 2 5
Seriousness class
2
1 0 1 52 0 2 5 3 0
Seriousness class
1
120
0
24
48
72
month
Predicted Monthly Admissions per 100000 population age 15-34
Based on a polynomial regression.
Figure 7: Trends in Prison Admissions per Capita in Georgia
A DDD methodology has the potential to improve the validity of inferences otherwise based on a DD
methodology, but not necessarily if we lack a principled basis for selecting one or more comparison
states. We turn to that issue next. We caution that the synthetic control methodology is emerging in
the evaluation literature.
4.4
Synthetic Control Methodology
We illustrate using California because we know California implemented a major reform
(Realignment) to reduce its prison population, and we might think that this reform would change the
ratio of seriousness class 1 offenders to the sum of seriousness class 1 and 2 offenders. Note that
California did not target its intervention to emphasize seriousness class 1 offenders over seriousness
class 2 offenders, although this seems like an interesting research question, and it nicely illustrates the
synthetic control methodology (Abadie, Diamond, & Hainmueller, 2010; Abadie, Diamond, &
Hainmueller, 2014). We do not claim, however, that this is a serious evaluation.
Still thinking about DDD, and still using the DD ratio, we face two problems: (1) What states should
be included in the comparison, and (2) What statistical test is appropriate for analysis? The synthetic
control methodology answers both questions.
First we treat the ratio SC 1/(SC 1 + SC 2) as the variable of interest. The left-hand panel of figure 8
shows that prior to the California intervention (the broken vertical line) this ratio had been decreasing
steadily and that after the intervention the ratio fell precipitously. From a DD perspective, this is
fairly strong evidence that Realignment has worked to reduce the proportion of offenders in
California prisons for relatively minor crimes (SC 1). The broken line shows the trend for the
Abt Associates
The NCRP Data as a Research Platform ▌pg. 22
synthetic cohort, identified as a cluster of states that experienced trends much like those experienced
in California prior to the intervention.16 After the intervention, the trend in the synthetic cohort states
continued its fairly linear pattern. The fact that the post-intervention trend in California departs from
the post-intervention trend in the synthetic cohort suggests that California successfully reduced the
proportion of SC 1 offenders to the sum of SC 1 and SC 2 offenders. We have used a DDD
perspective to strengthen the evidence from the DD perspective. However, we have not yet provided a
statistical test.
- .0 6
California
60
80
100
120
month
treated unit
synthetic control unit
- .1
- .0 8
California
.5
.5 5
r a ti o 1
.6
C a li fo r n ia
- .0 4
- .0 2
.6 5
0
.7
To understand the statistical test, first perform a mental calculation. Looking at the left-hand panel,
subtract the ratio for California from the ratio for the synthetic cohort. Graph that difference into the
panel on the right. Prior to the intervention, the difference is near zero, so the line on the right-hand
panel is flat until the intervention. Thereafter the line becomes increasingly negative.
60
80
100
120
m onth
Figure 8: California and Synthetic Cohort: SC 1/(SC 1 + SC 2)
Next, translate the right-hand panel from figure 8 to become the left-hand panel in figure 9. Then,
repeat the exercise applied to California to every other state; for each state imagine the counterpart to
the left-hand panel, and draw that imagined counterpart into the right-hand panel.17 The right-hand
panel looks like a ball of yarn, but what is important is that the trend for California forms a lower
boundary for the cluster of lines. Thirty states entered the analysis, so by chance, under the null
hypothesis of no treatment effect, California would provide the lower boundary in this figure with a
16
We refer readers elsewhere (Abadie, Diamond, & Hainmueller, 2010; Abadie, Diamond, & Hainmueller,
2014) for a technical explanation of identifying the synthetic control group. Intuitively the synthetic control
group comprises other states that have trends similar to those experienced in California and have
explanatory variables (such as arrests per capita) that have similar values. Members of the synthetic control
group are weighted by relevance so some states receive higher weights than do others. Many states receive
a weight of zero, meaning they are excluded from the synthetic control group.
17
Some states are so unique that they lack a synthetic cohort. They are excluded from the analysis so that the
33 states that entered the original analysis have been reduced to 30 states.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 23
0
California
- .0 5
Ca lifornia
- .1
- .1
- .0 8
- .0 6
C a li fo r n ia
- .0 4
.0 5
- .0 2
0
.1
probability of 1/30. Thus we reject the null of no treatment effect in California with a probability of
1/30 = 0.0333.
60
80
100
m onth
120
60
80
100
120
month
Figure 9: Test Statistic for Trend in SC 1/(SC 1 + SC 2)
Figure 10 provides a different measure for examining the trend: the ratio of SC 1 to the sum of SC 1,
SC 2 and SC 3. The impression is not much changed. From a DD perspective, illustrated by the lefthand panel, we have strong evidence that California’s Realignment has alter the composition of its
prison population in the intended direction. California has decreased the ratio of the least serious
offenders (as judged by offense seriousness) relative to other low seriousness offenders. From the
right-hand panel, we have evidence that this change in not spurious, because it has not occurred in
other states.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 24
.0 5
0
0
- .0 2
C a li fo r n ia
- .0 5
California
- .1
- .0 6
- .0 4
California
60
80
100
m onth
120
60
80
100
120
month
Figure 10: Test Statistics for Trend in SC 1/(SC 1 + SC 2 + SC 3)
The next figure is the counterpart to figure 10 but the ratio represented is SC 5/(SC 3 + SC 4 + SC 5).
This tests the null that California has increased the proportion of the most serious offense classes
relative to other relatively serious offense classes. One other state actually forms the upper boundary
for the ball of twine, so the effect is statistically significant at 0.066.
Another possible null is that California simultaneously decreased the use of incarceration for SC 1
relative to SC 1 and SC 2 and increased the use of incarceration for SC 5 relative to SC 4 and SC 5.
What we find is that California has reduced the use of prison for SC 1 (compared with SC 1 plus SC
2) by more than any other state and California has simultaneously increased the use of prison for SC 5
(compared with SC 4 plus SC 5) by more than every other state except one. These two trends are
independent, so it is highly unlikely that California could have accomplished these simultaneous
changes by chance.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 25
.0 6
.0 4
.0 4
.0 3
.0 2
California
- .0 4
- .0 1
0
- .0 2
0
.0 1
C a li fo r n ia
.0 2
California
60
80
100
m onth
120
60
80
100
120
month
Figure 11: Test Statistics for Trends in SC 5/(SC 3 + SC 4 + SC5)
Although we have rejected an interrupted time-series as a useful evaluation design, we have not
reached a conclusion that a DDD is superior to a DD. The synthetic cohort approach is recent and we
feel uncomfortable that not enough experience has accumulated to adopt the synthetic control method
as the principal evaluation method for statewide prison interventions. Furthermore, when it has been
applied, the synthetic control approach has assumed that one state (California, above) has
implemented the intervention while other states have not. Authorities have suggested how to deal
with multiple states (Abadie, Diamond, & Hainmueller, 2010), but when several states have adopted
an intervention—the JRI problem—the usefulness of the suggestions is not so obvious.
Our recommendation is using the DD and DDD in conjunction to strengthen conclusions in the face
of potential validity challenges. Because prison reforms typically target a specific prison population
defined by offense type (seriousness) and offender type (criminal history), it is practical to identify a
within-state counterfactual of offenses that are slightly less serious (hence not a target for the
intervention) and offenses that are slightly more serious (and hence not a target for the intervention).
Many evaluators would argue that this is a relatively strong basis for estimating a treatment effect
provided that SUTVA is met.
Nevertheless, a known deficiency of a DD framework is that pre-intervention trends may not portend
post-intervention trends in the absence of the intervention. Using the logic of an interrupted timeseries framework by limiting the bandwidth is helpful for dealing with this validity challenge, but we
have suggested another procedure, namely using a DDD framework to test whether the trends in the
state being evaluated differ substantially from the trends in other states. Not all states may offer good
comparisons, so using statistical tests, we have applied the synthetic estimation framework to select
states.
If the DD-estimated effect is not statistically significant or substantively meaningful, we probably halt
the investigation. If it is significant/substantively meaningful, we then apply DDD through the
synthetic estimation approach. However, it seems unreasonably conservative to put heavily reliance
on statistical significance from the synthetic comparison approach. After all, if we require both tests
Abt Associates
The NCRP Data as a Research Platform ▌pg. 26
(the DD and the DDD), under the null the probability of rejecting the null is no longer 0.05, but
rather, 0.05x0.05 = 0.0025. Clearly the test is too conservative.
We are unsure of an optimal test, but it seems sensible to mix quantitative and qualitative tests. The
quantitative test is based on the DD. As already noted, if we fail to reject the null, then testing ceases.
If we reject the null, the qualitative test is based on the DDD. To pass the qualitative test, we would
expect California to fall near the lower or upper envelope of the multiple curves, but requiring
California to form the envelope (the only way to achieve p < 0.05 given 30 states in the study) seems
too severe.
5.
Conclusions
This paper has discussed approaches to evaluating state-level reforms intended to reduce the use of
prison for selected classes of offenders. Evaluation is difficult because random assignment is
impractical and evaluation requires other approaches. Alternative approaches face validity and
reliability challenges because it is difficult to identify suitable counterfactuals, and when they are
identified, sample sizes are small.
We believe that interrupted time-series are poor designs that can lead to spurious findings, sometimes
causing evaluators to reject interventions that are beneficial and sometimes causing evaluators to
accept interventions that are ineffective. When the intervention targets a class of offenders, then a
class of similar offenders within the same state may be a suitable counterfactual. This is the logic of a
difference-in-difference design. Some additional rigor may be gained by augmenting the differencein-differences with a difference-in-difference-in-differences approach, comparing trends across states.
The problem is to identify suitable states for comparison and to identify statistical tests that recognize
the small sample involved in the comparison. Synthetic control may provide a useful approach.
We have skirted or only briefly mentioned important issues. One issue is identifying the
counterfactuals. We based the counterfactuals exclusively on five offense seriousness classes, but this
is probably inadequate for many evaluation questions. As already mentioned, most prison reforms
target certain offenses and offenders, and the counterfactual should be built around those types.
Another issue is that states use their prisons in different capacities. For example, some states may
frequently send offenders convicted of domestic assault to prison; other states may do so rarely. If
offenders convicted of domestic assault are not part of the targeted group, it seems inappropriate to
include them in any analysis that makes cross-state comparisons. This is just to say that an evaluator
must think carefully about appropriate counterfactuals, and the choice of a counterfactual will hang
on the evaluation question.
Especially when drawing cross-state comparisons, an evaluator needs to consider what other
interventions are occurring. For convenience, our illustrations assume that no other interventions were
occurring, but that assumption was for convenience and a serious evaluation would carefully
determine its truth. California was an extreme choice; no other state, to our knowledge, has imposed
such a strong change on its prison system during the same time frame. This will not always be the
case, however. We were motivated to think about this problem because of the Justice Reinvestment
Initiative, which is being implemented by several states, in different forms, simultaneously. From an
evaluation standpoint, it may not make sense to estimate the size of the treatment effect by comparing
JRI participants in the synthetic control framework. The larger concern is that two or more states may
Abt Associates
The NCRP Data as a Research Platform ▌pg. 27
simultaneously implement interventions, and while it may be useful to understand whether one type
of intervention is preferable to another, the larger question regards what each accomplishes. From the
DDD perspective, this requires comparing a state (or states) that implemented interventions to states
that did not. In turn, this requires detailed knowledge of what states have done to moderate their
prison populations. The NCRP program assembles useful data (known as the fact sheets), providing
some basis for selection.
Finally, the discussion has concerned prison population composition, but this is not the only type of
question that might be posed and answered using NCRP data. The NCRP is especially useful for
studying recidivism defined as returning to prison in the same state. (The NCRP team is working on
linking NCRP data across states so over time the definition will be expanded.) Questions about
recidivism are equally amenable to the research designs posed here.
Our expectation is to apply the recommended approach to the Justice Reinvestment Initiative. JRI was
implemented between 2010 and 2013 so with 2014 data we should be able to assess the impact of JRI
interventions.
References
Abadie, A., Diamond, A., & Hainmueller, J. (2010). Synthetic Control Methods for Comparative
Case Studies Estimating the Effect of California's Tobacco Control Program. Journal of the
American Statistical Association, vol 105 (490), 493-506.
Abadie, A., Diamond, A., & Hainmueller, J. (2014). Comparative Policits and the Synthetic Control
Method. American Journal of Political Science, pp: 59(2), 495-510.
Angrist, J., & Pischke, J. (2009). Mostly Harmless Econometrics: An Empiricist's Companion.
Princeton: Princeton University Press.
Berk, R. (2005). Randomized Experiments as the Bronze Standard. Journal of Experimental
Criminology, 1(4), pp. 417-433.
Borenstein, M., Hedges, L., Higgiens, J., & Rothstein, H. (2010). A Basic Introduction to FixedEffect and Random-Effects Models for Meta-Analysis. Research Synthesis Methods, 97-111.
Borenstein, M., Hedges, L., Higgins, J., & Rothstein, H. (2009). Introduction to Meta-Analysis. West
Sussex, United Kingdom: John Wiley & Sons, Inc.
Cameron, A., & Trivedi, P. (2005). Microeconometrics: Methods and Applications. Cambridge, UK:
Cambridge University Press.
Imbens, G., & Lemieux, T. (2007). Regression Discontinuity Designs: A Guide to Practice.
Cambridge, MA: National Bureau of Economic Research Working Paper.
Imbens, G., & Rubin, D. (2015). Causal Inference for Statistics, Social, and Biomedical Sciences: An
Introduction. Cambridge, MA: Cambridge University Press.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 28
Imbens, G., & Wooldridge, J. (2009). Recent Developments in the Econometrics of Program
Evaluation. Journal of Economic Literature, 47(1).
Lee, M. (2005). Micro-Econometrics for Policy, Program and Treatment Effects. Oxford, UK:
Oxford University Press.
Luallen, J., Rhodes, W., Gaes, G., Kling, R., & Rich, T. (2014). A Description of Computing Code
Used to Identify Correctional Terms and Histories. . Cambridge, MA: Abt Associates Inc.
Manski, C. (2007). Identification for Prediction and Decision. Cambridge, MA: Harvard University
Press.
Morgan, S., & Winship, C. (2015). Counterfactuals and Causal Inference: Methods and Principals
for Social Research, 2nd Edition. Cambridge, MA: Cambridge University Press.
Orr, L. (1999). Social Experiments: Evaluating Public Policy Progrms with Experimental Methods.
Thousand Oaks, California: Sage Publications.
Rhodes, W. (2012). Meta-Analysis: An Introduction using Regression Models. Evaluation Review
vol. 36 (1), 24-71.
Rhodes, W., & Jalbert, S. (2013). Regression Discontiuity Design in Criminal Justice Evaluation: An
Introduction and Illustration. Evaluation Review, 239-273.
Rosenbaum, P. (2009). Design of Observational Studies. New York: Springer.
Wooldridge, J. (2007). What's New In Econometrics? Lecture 10 Difference-in-Differences
Estimation.
Abt Associates
The NCRP Data as a Research Platform ▌pg. 29
Appendix C
Results of state survey on their ability to provide prisoners’ 9‐digit social security number to NCRP
Yes
Arkansas
Connecticut
Hawaii
Kansas
Louisiana
Michigan
New Hampshire
New Jersey
New Mexico
North Dakota
South Dakota
Vermont
West Virginia
Wyoming
Probably/Likely/Very likely
Arizona
California
Indiana
Iowa
Kentucky
Maine
Massachusetts DOC
Massachusetts parole
Mississippi
Missouri
New York
Ohio
Oklahoma
Pennsylvania parole
Rhode Island
South Carolina DOC
South Carolina parole
Texas
Utah
Washington
Maybe
Colorado
Delaware
Georgia DOC
Minnesota
North Carolina
Unlikely/Very unlikely
Alaska (law restricts access)
Idaho
Maryland
Nebraska
Nevada parole
Wisconsin
No
DC (CSOSA)
Georgia parole (by law)
Nevada DOC (policy)
Oregon (policy)
Pennsylvania DOC (policy)
Tennessee
Don't know Did not respond
Alabama
Florida
Montana
Illinois
Virginia
Appendix D
Phone script to introduce reporting year 2015 data collection for states already submitting data to
NCRP
NCRP 2015 Data Collection Protocol and Interview Guide – CURRENTLY CONTRIBUTING
STATES
Prior to initial conversation with state:
Get background information:
o Review prior conversations with state (to re-familiarize yourself)
o Get 2014 submission date (to identify target date for 2015 data)
o Find out the NPS and APS contacts for the state
o Look up what we are thanking them for (see NCRP points of contact.xls)
o review Fact Sheet, to re-familiarize yourself with the state
o Review data quality issues Jeremy and Ryan identified (see K:\Projects\NCRP\State
Folders)
o Significant, unexplained differences between NCRP and NPS or DOC annual report
Determine what we need to ask them for
o 2015 NCRP data
o Other NCRP data: D records, ABD from prior years, additional ABD data elements, EF
records
o Approve/review Fact Sheet
Email the primary contact to set up a time to talk. The purpose of the call will be to:
o Talk about the 2015 data request.
o Talk about improvements we have made to NCRP
o Get your ideas on other improvements we can make to NCRP
Record initial and follow-up attempts to reach POC on your tracking sheet.
General outline conversation with primary point of contact (will vary depending on your relationship
with the POC and the POC’s familiarity with NCRP and our project)
Confirm this is a good time to talk
Thank them for what they did in 2015 (2014 data request, have list ready)
Indicate we have made a number of improvements to NCRP over the past year. For example:
o Construction of EF term records
o More states are participating
o New data quality controls; ABD term records
Discuss the 2015 data request
o Highlight what has changed for 2015 – request 9-digit SSN, address of last residence prior
to imprisonment, security level at which prisoner is being held
o Confirm that they’ll be able to submit 2015 data (e.g., not transitioning to a new system)
Set a target date for submission, based on what they did last year
o If a new state, indicate that the includes and excludes are in the FAQ.
In addition to getting 2015 data, we have other initiatives planned for 2016 and beyond that will
improve the NCRP
o [as appropriate] data linkage to external federal sources if permitted by the states
o [as appropriate] want to fill in the holes from prior years – see if they can submit old data
o [as appropriate] want to get additional data elements – see if they can do this
o [as appropriate] want to get EF records – see if they can submit EF records for the first
time. Mention that BJS grant funding may be available, and see if they are interested.
1
o [as appropriate] want to re-design of how NCRP does parole/PCCS – as appropriate, ask
parole/PCCS questions, identify who to talk with, or set up time to talk with the POC about
this
Ask them if they have any questions about NCRP.
Confirm next steps for them (e.g., submit 2015 data, then submit old data)
Indicate what you will be sending to them (see list below)
o Ask them if they prefer the materials mailed or emailed
o Ask them what pieces of the mailing should be sent to others (above or below them) –
confirm contact information. Ask about other persons in our state contact list.
Thank them for participating in NCRP
After the initial conversation
Record the date of the conversation on your tracking sheet
Within 2 days of the call, prepare the materials to be mailed (or emailed) to them
o Data request materials
BJS cover letter: add POC name and address, save in K:\Projects\NCRP\2015 Data
Collection\Materials sent to states\[state name], print on BJS letterhead
Abt cover letter: add POC name and address, customize depending on what they
did last year and what they agreed to do in 2016, save in K:\Projects\NCRP\2015
Data Collection\Materials sent to states\[state name], print on Abt letterhead
FAQ
o Use hand-addressed Abt (9x12) envelope, if mailing
Tom reviews the materials before sending
Mail (1st class) or email
Record date of mailing on your tracking sheet
Send NCRP Newsletter to other state contacts, as appropriate
If data is not received by March 31, 2016
Check tracking sheet to see whether they previously said they would be late submitting data (e.g.,
because of legislative session work)
Email point of contact (customize this email depending on your relationship with the contact):
o “We are checking back with you on the status of our request for 2015 NCRP data. You
had earlier indicated to us that you would be able to submit these data by March 31, 2016.
Please contact me if you have any questions. Thank you again for participating in NCRP.”
Record email on tracking sheet
Follow-up if no response in 2-3 days
Record response on tracking sheet
2
Appendix E
Phone script to introduce reporting year 2015 data collection for states that have not recently
submitted data to NCRP
NCRP 2015 Data Collection Protocol and Interview Guide –LAPSED STATES OR THOSE
NOT CURRENTLY SUBMITTING
Prior to initial conversation with state:
Get background information:
o Review any prior conversations with state (to re-familiarize yourself)
o If state has submitted in the past, review what parts of NCRP and what years were
provided.
o Find out the NPS and APS contacts for the state
o review Fact Sheet, to re-familiarize yourself with the state
o If the state has submitted in the past, review any data quality issues that Jeremy and Ryan
have identified (see K:\Projects\NCRP\State Folders)
o Significant, unexplained differences between NCRP and NPS or DOC annual report
o For states that have never submitted, consider having BJS send a letter to the DOC
commissioner to solicit participation.
Determine what we need to ask them for
o 2015 NCRP data
o Other NCRP data: D records, ABD from prior years, additional ABD data elements, EF
records
Email the primary contact to set up a time to talk. The purpose of the call will be to:
o Talk about the 2015 data request and what’s new for this year.
o Talk about advantages of submitting to NCRP (access to NCRP website for state to state
analytic tool, use of NCRP data by outside groups for research)
o Discuss the reduced list of variables we are requesting for lapsed and new states.
o Mention that if budget constraints prevent the DOC from doing the programming required
to extract data for NCRP, BJS does make small one-time grants to assist states that have
never submitted, have lapsed in submission, or are making IT system changes that require
reprogramming extraction code.
o Get your ideas on other improvements we can make to NCRP (lapsed states only)
Record initial and follow-up attempts to reach POC on your tracking sheet.
General outline conversation with primary point of contact (will vary depending the POC’s
familiarity with NCRP and our project)
Confirm this is a good time to talk
o For states that have never submitted, confirm that they are the person you should be talking
to (they can authorize participation in NCRP). Also ask whether we need to contact a
separate person for the parole records.
[If state has previously submitted] Thank them for past participation, and make the case for
restarting submission (new states have come on, we will accept reduced variable list, etc)
Thank them for taking the time to speak with you, introduce the NCRP and its many uses by
federal, state, nonprofit, and academic researchers.
o Discuss how important an administrative data collection is to BJS, since they can only get
out to field the survey of prison inmates every 7-10 years.
o Stress that once the extraction program has been set up, only very minor changes need to
be made in subsequent years to provide annual data.
1
o Mention that we will accept a reduced list of variables to get them started, and that BJS can
provide one-time small grants to support programming of extraction programs.
Indicate BJS is committed to NCRP and Abt Associates, their data collection agent, has made a
number of improvements to NCRP over the past 5 years For example:
o Construction of term records for both prison and PCCS records
o More states are participating
o New data quality controls
o Annual data providers meeting
Discuss the 2015 data request
o Highlight what has changed for 2015 – request 9-digit SSN, address of last residence prior
to imprisonment, security level at which prisoner is being held
o Confirm that they’ll be able to submit 2015 data (e.g., not transitioning to a new system)
Set a target date for submission, based on what they did last year
o If a new state, indicate that the includes and excludes are in the FAQ.
In addition to getting 2015 data, we have other initiatives planned for 2016 and beyond that will
improve the NCRP
o [as appropriate] data linkage to external federal sources if permitted by the states
o [as appropriate] want to fill in the holes from prior years – see if they can submit old data
at the same time as the 2015 data (should just require a change in the year in the extraction
program)
o [as appropriate] want to get E,F records – see if they can submit E,F records for the first
time. Mention that BJS grant funding may be available, and see if they are interested.
o [as appropriate] ask parole/PCCS questions, identify who to talk with, or set up time to talk
with the POC about this
Ask them if they have any questions about NCRP.
Confirm next steps for them (e.g., submit 2015 data, then submit old data)
Indicate what you will be sending to them (see list below)
o Ask them if they prefer the materials mailed or emailed
o Ask them what pieces of the mailing should be sent to others (above or below them) –
confirm contact information. Ask about other persons in our state contact list.
Thank them for participating in NCRP
After the initial conversation
Record the date of the conversation on your tracking sheet
Within 2 days of the call, prepare the materials to be mailed (or emailed) to them
o Data request materials
BJS cover letter: add POC name and address, save in K:\Projects\NCRP\2015 Data
Collection\Materials sent to states\[state name], print on BJS letterhead
Abt cover letter: add POC name and address, customize depending on what they
have agreed to do in 2016, save in K:\Projects\NCRP\2015 Data
Collection\Materials sent to states\[state name], print on Abt letterhead
FAQ
o Use hand-addressed Abt (9x12) envelope, if mailing
Tom reviews the materials before sending
Mail (1st class) or email
Record date of mailing on your tracking sheet
Send NCRP Newsletter to other state contacts, as appropriate
2
If data are not received by March 31, 2016
Check tracking sheet to see whether they previously said they would be late submitting data (e.g.,
because of legislative session work)
Email point of contact (customize this email depending on your relationship with the contact):
o “We are checking back with you on the status of our request for 2015 NCRP data. You
had earlier indicated to us that you would be able to submit these data by March 31, 2016.
Please contact me if you have any questions. Thank you again for participating in NCRP.”
Record email on tracking sheet
Follow-up if no response in 2-3 days
Record response on tracking sheet
3
Appendix F
Introductory letter from BJS to data respondents for collection of 2015 NCRP data
U.S. Department of Justice
Office of Justice Programs
Bureau of Justice Statistics
Washington, D.C. 20531
DATE
Name
Agency
Address
City
State, zip
Dear_____:
We are writing to request your participation in the National Corrections Reporting Program
(NCRP). Data are now being collected for the 2015 reporting year by Abt Associates Inc., our
data collection agent.
Last year all 50 states submitted at least some NCRP data. We are confident that in 2016 we will
have 100% participation. For 2015, our emphasis will be on increasing the number of states that
submit key offender identifiers (State ID and FBI number) and post-confinement community
supervision admission (Part E) and release (Part F) records. Please note that there are no new
variables or records in this year’s request.
As provided 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. The NCRP collection underwent its 3-year
clearance review by the Office of Management and Budget in 2012 and was approved; you can
read the application and review comments at:
http://www.reginfo.gov/public/do/PRAViewICR?ref_nbr=201208-1121-005.
Finally, we want to alert you that in addition to this request for NCRP data, if you are the
respondent for other annual BJS data collections, you will receive separate cover letters for these
collections, including the National Prisoner Statistics (NPS), Annual Probation and Parole
Surveys (APS), Capital Punishment, and Deaths in Custody Reporting Program (DCRP). We
appreciate the amount of time and energy that you expend in providing us these data. Without
your assistance, BJS would be unable to provide comprehensive and accurate statistics on the
correctional populations in the United States.
On behalf of BJS, Abt will be in contact with your agency shortly to launch the 2015 data
collection process. In the meantime, if you have any questions please feel free to contact the Abt
Project Director, Tom Rich, at 617-349-2753 or Tom_Rich@abtassoc.com or the BJS Program
Manager, Ann Carson, at 202-616-3496 or elizabeth.carson@ojp.usdoj.gov. Once again, many
thanks for your participation in BJS’ NCRP program.
Sincerely,
William J. Sabol, Ph.D.
Director
Bureau of Justice Statistics
E. Ann Carson, Ph.D.
Statistician and Program Manager, NCRP
Bureau of Justice Statistics
Appendix G
Introductory letter from data collection agent to data respondents for collection of 2015 NCRP data
January 12, 2016
,
Dear :
On behalf of the Bureau of Justice Statistics (BJS), I want to thank you for participating in the
National Corrections Reporting Program (NCRP). Last year all 50 states submitted at least some
NCRP data. We are confident that in 2015 we will continue to have 100% participation.
For this year’s request, we are requesting that you submit 2015 Parts A, B, D, E, and F – the same
as you have done in the past. Data request instructions and submission procedures are attached.
If possible, we would appreciate receiving these data by March 31, 2016.
BJS has obtained permission from the Office of Management and Budget through its clearance
procedure to request variables that will allow BJS and other researchers to better characterize the
geographic and security profile of offenders, as well as to link the NCRP data to other federal
datasets if permitted by . If possible, please add the following
items to your submission:
Add 9-digit social security number and address of last known residence prior to
imprisonment to the NCRP data files you currently submit.
Add the security level of custody for each inmate in NCRP Parts A and D (prison
admission and yearend custody records).
If you have any questions about NCRP or this data request, please contact me at 617-349-2753 or
tom_rich@abtassoc.com. Again, we appreciate your support of NCRP.
Sincerely,
Tom Rich
NCRP Site Liaison
55 Wheeler Street
Cambridge, MA 02138
Office 617.492.7100
abtassociates.com
Appendix H
Instructions for NCRP data submission, reporting year 2015
2015 NCRP
Data Request
Instructions
Prison and Post
Confinement
Community
Supervision Records
(Parts A, B, D, E, and
F)
January 2016
Contacts:
Tom Rich
NCRP Project Director
and Site Liaison
617-349-2753
tom_rich@abtassoc.com
Michael Shively
NCRP Site Liaison
617-520-3562
michael_shively@abtassoc.com
Abt Associates Inc.
55 Wheeler Street
Cambridge, MA 02138
2015 NCRP Data Request Instructions
Table of Contents
Overview................................................................................................................................................ 1
What’s New for 2015 ....................................................................................................................... 1
General Data Submission Instructions ............................................................................................. 2
Part A (Prison Admissions) Instructions ............................................................................................ 4
Part B (Prison Releases) Instructions ................................................................................................. 8
Part D (Prison Custody) Instructions ............................................................................................... 13
Part E (Post Confinement Community Supervision Admissions) Instructions ............................ 17
Part F (Post-Confinement Community Supervision Releases) Instructions ................................. 21
Appendix.
Additional Information on NCRP Variables .................................................... 25
Variable 1: County in Which Sentence Was Imposed ................................................................... 25
Variable 2: Inmate ID Number ...................................................................................................... 25
Variable 3: Date of Birth................................................................................................................ 26
Variable 4: Sex............................................................................................................................... 26
Variable 5: Race............................................................................................................................. 26
Variable 6: Hispanic Origin ........................................................................................................... 27
Variable 7: Highest Grade Completed ........................................................................................... 28
Variable 8: Date of Admission to Prison ....................................................................................... 29
Variable 9: Type of Admission to Prison....................................................................................... 30
Variable 10: Jurisdiction on Date of Admission ............................................................................ 34
Variable 11: Prior Jail Time ........................................................................................................... 35
Variable 12: Prior Prison Time ...................................................................................................... 35
Variable 13: Offenses .................................................................................................................... 36
Variable 14a: Offense with Longest Maximum Sentence ............................................................. 37
Variable 14b: Sentence Length for Variable 14a Offense ............................................................. 37
Variable 15: Total Maximum Sentence Length ............................................................................. 38
Abt Associates Inc.
Contents ▌pg. i
Variable 17: Location Where Inmate is to Serve Sentence............................................................ 40
Variable 18: Additional Offenses since Admission Date .............................................................. 41
Variable 19: Additional Sentence Time since Admission ............................................................. 41
Variable 20: Prior Felony Incarcerations ....................................................................................... 42
Variable 21: AWOL or Escape ...................................................................................................... 43
Variable 22a: Community Release Prior to Prison Release ........................................................... 43
Variable 22b: Number of Days on Community Release................................................................ 44
Variable 23a: Date of Release from Prison .................................................................................... 44
Variable 23b: Location at Time of Prison Release ........................................................................ 44
Variable 24: Agencies Assuming Custody at Time of Prison Release .......................................... 45
Variable 25: Type of Release From Prison .................................................................................... 46
Variable 26: Date of Release from Post Confinement Community Supervision ........................... 49
Variable 27: Type of Release from Post Confinement Community Supervision .......................... 50
Variable 28: Supervision Status Just Prior to Release ................................................................... 53
Variable 30: Inmate State ID Number ........................................................................................... 54
Variable 31a: Indeterminate Sentence ........................................................................................... 54
Variable 31b: Determinate Sentence.............................................................................................. 55
Variable 31c: Mandatory Minimum Sentence ............................................................................... 56
Variable 31d: Truth in Sentencing Restriction .............................................................................. 56
Variable 32: Length of Court-Imposed Sentence to Community Supervision .............................. 57
Variable 33: Parole Hearing / Eligibility Date ............................................................................... 57
Variable 34: Projected Release Date.............................................................................................. 58
Variable 35: Mandatory Release Date ........................................................................................... 59
Variable 36: First Name ................................................................................................................. 60
Variable 37: Last Name ................................................................................................................. 60
Variable 38: Facility Name ............................................................................................................ 61
Variable 39: FBI Number .............................................................................................................. 61
Variable 40: Prior Military Service................................................................................................ 61
Abt Associates Inc.
Contents ▌pg. ii
Variable 41: Date of Last Military Discharge ................................................................................ 62
Variable 42: Type of Last Military Discharge ............................................................................... 62
Variable 43: Date of Admission to Post Confinement Community Supervision ........................... 63
Variable 44: Type of Admission to Post Confinement Community Supervision .......................... 63
Variable 45: County Where Offender was Released / County Where PCCS Office is Located ... 64
Variable 46: Social Security Number………………………..…………………………………...65
Variable 47: Street Address of Residence Prior to Imprisonment………………………………..65
Variable 48: City of Residence Prior to Imprisonment…………………………………………..65
Variable 49: State of Residence Prior to Imprisonment…………………...……………………..66
Variable 50: Zip Code of Residence Prior to Imprisonment……………………………………..66
Variable 51: Custodial Security Level at which Inmate is Imprisoned…………………………..66
Abt Associates Inc.
Contents ▌pg. iii
Overview
The National Corrections Reporting Program (NCRP) collects offender-level information from state
departments of correction and community supervision on admissions to and releases from prisons and
post confinement community supervision programs. Abt Associates is the NCRP data collection
agent for the Bureau of Justice Statistics, the federal agency that administers NCRP. BJS has
administered NCRP since 1983. Contact your NCRP site liaison (Tom Rich, at
tom_rich@abtassoc.com or 617-349-2753 or Mike Shively, at michael_shively@abtassoc.com or
617-520-3562) for more information. Or visit the NCRP website at www.ncrp.info.
For 2015, states are asked to submit three prison files:
Prison Admissions (Part A): one record for each admission of a sentenced offender to the state’s
prison system during calendar year 2015.
Prison Releases (Part B): one record for each release of a sentenced offender from the state’s
prison system during calendar year 2015.
Prison Custody (Part D): one record for each sentenced offender in the physical custody of the
state’s prison system on December 31, 2015.
For 2015, states are also asked to submit two post-confinement community supervision (PCCS)
files:
Post Confinement Community Supervision1 Admissions (Part E): one record for each admission
to a post-confinement community supervision program.
Post Confinement Community Supervision Releases (Part F): one record for each release from a
post-confinement community supervision program.
The detailed instructions below for Parts A, B, D, E, and F include the NCRP definitions of
admissions, releases, and other terminology. The NCRP definitions may vary from the definitions
your state uses.
What’s New for 2015
BJS has obtained clearance from OMB (1121-005) to collect the following pieces of information for
all parts (A – F):
9-digit social security number
BJS has obtained clearance from OMB (1121-005) to collect the following pieces of information for
prison and parole admission records (A, E):
1
Post Confinement Community Supervision means sentenced offenders serving a period of community
supervision immediately after release from prison.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 1
Address of last residence prior to imprisonment, consisting of street address, city, state, and
zip code
In addition, BJS has added the following item to prison yearend custody records (Part D):
Security level at which the inmate is imprisoned
General Data Submission Instructions
Is there a required format or coding scheme for the data?
There is no required format or file type for the data you submit; use whatever is most convenient
for you.
There is no required set of codes for the categorical NCRP variables (e.g., race, prison admission
type). The documentation in this manual includes suggested “NCRP format” codes, but you can
use whatever internal codes your agency uses. As necessary, Abt will re-code your internal
agency codes into the standardized NCRP codes.
What if I am unable to provide all the requested data?
If your agency does not collect one or more of the requested data elements or providing them
would be an excessive burden (or is not allowed under agency policy), those data elements do not
have to be included in the data submission. The instructions for each Part also highlight the
“core” data elements that are most important to NCRP.
When is the data submission due?
The target date for submitting NCRP data is March 31st, but we understand that agency
constraints in many states preclude meeting that target date. The Abt site liaison will work with
each state to set a realistic target date.
How do I send the data to Abt Associates?
The preferred method for submitting data is via the NCRP data transfer site
(transfer.abtassoc.com). This site is compliant with FIPS (Federal Information Processing
Standard) 140-2 and meets all the requirements of the Federal Information Security Management
Act (FISMA) and the Privacy Act. The data are automatically encrypted during transit.
When you are ready to submit data, contact your NCRP site liaison2 to obtain a unique username
and password for the transfer portal, or to make other submission arrangements. Please protect
your transfer portal username and password. Instructions on how to use the transfer site are
available from your Abt site liaison.
What happens after we submit data?
2
Tom Rich, at tom_rich@abtassoc.com or 617-349-2753, or Mike Shively, at
michael_shively@abtassoc.com or 617-520-3562
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 2
Abt will verify the contents of the data files and conduct a series of validity checks on the data
(including comparing the submitted data to your submissions from prior years). Typically, this
will be accomplished within 2-4 weeks of receipt of your data. Your Abt site liaison will then
contact you to review the findings. Having a thorough understanding of what data you submit is
necessary in order to construct valid and reliable national NCRP datasets.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 3
Part A (Prison Admissions) Instructions
The data file you produce for Part A should contain one data record for each admission of a
sentenced inmate to your prison system during 2015, regardless of sentence length or
jurisdiction.
NCRP defines admissions as including:
new court commitments;
revocations from probation, parole, or other types of post-confinement community supervision;
transfers from other jurisdictions;
escape or AWOL returns;
returns from appeal or bond.
Include in Part A:
Admissions of sentenced inmates to your prison facilities.3
Admissions of sentenced inmates under your jurisdiction to county or local jails.
Admissions of sentenced inmates under your jurisdiction to in-state private prisons, including
both privately owned facilities and facilities operated by a private entity under contract to the
state.
Exclude from Part A:
Admissions of sentenced inmates to one of your prison facilities who are being transferred from
another one of your prison facilities.
Inmates re-entering a prison facility after a temporary leave of 30 days or less (e.g., for a court
appearance, funeral furlough, or medical care).
Admissions of sentenced inmates under your jurisdiction to Federal facilities, another state’s
facilities, or out-of-state private facilities.
Admissions of unsentenced inmates to your prison facilities (e.g., inmates awaiting trial, civil
commitments)
The variables requested in the Part A data set are listed on the next page. Most of these variables are
also in the Part B and D requests. Refer to the Appendix for additional information on these variables.
3
Prison facilities include prisons, penitentiaries, and correctional institutions; boot camps; prison farms;
reception, diagnostic, and classification centers; release centers, halfway houses, and road camps; forestry
and conservation camps; vocational training facilities; prison hospitals; and drug and alcohol treatment
facilities for prisoners. For inmates under home confinement, a private residence is not considered a prison
facility.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 4
The Part A (Prison Admissions) variables are listed below in the table. If you have limited resources for responding to this data request, please
focus on the core variables. Additional information on the variables is in the Appendix.
Category
Offender
#
2
Name
Inmate ID Number
30
State ID Number
39
FBI Number
36
Definition
A unique number that identifies an offender within the agency for this
admission and all subsequent admissions.
The offender’s unique, fingerprint-supported state identification number
Core
Variable
First Name
The unique identification number given by the Federal Bureau of
Investigation/ Interstate Identification Index to each offender
The offender’s first name
37
Last Name
The offender’s last name
46
47
SSN
Residential Street
Address
Residential City
Residential State
Residential Zip
Code
Date of Birth
Sex
Race
Hispanic Origin
Highest Grade
Completed
Prior Military Service
Date of Last Military
Discharge
Type of Last Military
Discharge
9-digit social security number
Street address of residential address prior to imprisonment
City of residence prior to imprisonment
State of residence prior to imprisonment
Zip code of residence prior to imprisonment
The offender’s date of birth
The offender’s biological sex
The offender’s race
Is the offender of Hispanic origin?
The highest academic grade level the offender completed prior to
admission to prison on the current sentence
Did the offender ever served in the U.S. Armed Forces?
The date the offender was discharged from the U.S. Armed Forces for the
final time
The type of discharge the inmate received from the U.S. Armed Forces
48
49
50
3
4
5
6
7
40
41
42
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 5
Category
Sentence
#
1
11
Name
County in Which
Sentence was
Imposed
Prior Jail Time
12
Prior Prison Time
13
Offenses
14a
Offense with
Longest Maximum
Sentence
Sentence Length for
Variable 14a
Offense
Total Maximum
Sentence Length
Indeterminate
Sentence
Determinant
Sentence
Mandatory Minimum
Sentence
Truth in Sentencing
Law Restriction
14b
15
31a
31b
31c
31d
32
Prison
Admission
8
Abt Associates Inc.
Length of CourtImposed Sentence
to Community
Supervision
Date of Admission to
Prison
Definition
The county where the court imposing the current sentence is located
Core
Variable
The length of time served in jail prior to the date of admission (Variable 8)
and credited to prison service for the current sentence
The length of time served in prison prior to the date of admission (Variable
8) and credited to prison service for the current sentence
Crime(s) for which the offender was admitted to prison on the current
sentence(s), including the number of counts for each offense.
Of the crimes coded in Variable 13, the ONE crime for which the inmate
received the longest sentence
The maximum sentence as stated by the court that the offender is required
to serve for the offense listed in Variable 14a
The longest length of time as stated by the court that the offender could be
required to serve for all offenses specified in Variable 13 (Offenses)
Does the total maximum sentence (Variable 15) include an indeterminate
sentence?
Does the total maximum sentence (Variable 15) include a determinate
sentence?
Does the total maximum sentence (Variable 15) include a mandatory
minimum sentence?
Is the total maximum sentence (Variable 15) restricted by a Truth in
Sentencing Law mandating that a certain percentage of the court- imposed
sentence be served in prison?
The amount of time which the court states that the offender is required to
serve under community supervision after release from prison
The most recent date the offender was admitted into the custody of the
state prison system on the current sentence
2015 NCRP Data Request Instructions ▌pg. 6
Category
#
9
10
17
Anticipated
Release
from Prison
33
34
35
Abt Associates Inc.
Name
Type of Admission
to Prison
Jurisdiction on Date
of Admission
Location where
Offender is to Serve
Sentence
Parole
Hearing/Eligibility
Date
Projected Release
Date
Mandatory Release
Date
Definition
The reason an offender entered into the physical custody of a correctional
facility on the date provided in Variable 8 (Admission Date) of the current
record
The state with the legal authority to enforce the prison sentence
The type of facility in which the offender will be incarcerated to serve time
for his/her crime.
The date the offender is eligible for review by an administrative agency
such as a parole board, to determine whether he or she will be released
from prison
The projected date on which the offender will be released from prison
The date the offender by law must be conditionally released from prison
2015 NCRP Data Request Instructions ▌pg. 7
Core
Variable
Part B (Prison Releases) Instructions
The data file you produce for Part B should contain one data record for each release of a sentenced
inmate from your prison system during 2015, regardless of sentence length or jurisdiction.
NCRP defines releases as including:
conditional releases from prison to parole, probation, or other forms of post-confinement
community supervision;
unconditional releases;
releases or transfers to other authorities;
deaths;
releases on appeal or bond if credit for time served is not given while on release;
escapes from custody.
Include in Part B:
Releases of sentenced inmates from your prison facilities4, regardless of jurisdiction or sentence
length.
Releases of sentenced inmates under your jurisdiction from county or local jails.
Releases of sentenced inmates under your jurisdiction from in-state private prisons, including
both privately owned facilities and facilities operated by a private entity under contract to the
state.
Exclude from Part B:
Sentenced inmates who are being transferred from one of your facilities to another one of your
prison facilities.
Temporary releases of sentenced inmates of 30 days or less (e.g., for a court appearance, funeral
furlough, or medical care).
Releases of sentenced inmates under your jurisdiction from Federal facilities, another state’s
facilities, or out-of-state private facilities.
Releases of unsentenced inmates from your prison facilities (e.g., inmates awaiting trial, civil
commitments)
4
Prison facilities include prisons, penitentiaries, and correctional institutions; boot camps; prison farms;
reception, diagnostic, and classification centers; release centers, halfway houses, and road camps; forestry
and conservation camps; vocational training facilities; prison hospitals; and drug and alcohol treatment
facilities for prisoners. For inmates under home confinement, a private residence is not considered a prison
facility.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 8
The variables requested in the Part B data set are listed on the next page. Most of these variables are
also in the Part A and D requests. Refer to the Appendix for additional information on these variables.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 9
The Part B (Prison Releases) variables are listed below in the table. If you have limited resources for responding to this data request, please
focus on the core variables. Additional information on the variables is in the Appendix.
Category
Offender
#
2
Name
Inmate ID Number
30
39
State ID Number
FBI Number
36
37
46
3
4
5
6
7
11
First Name
Last Name
SSN
Date of Birth
Sex
Race
Hispanic Origin
Highest Grade
Completed
Prior Military Service
Date of Last Military
Discharge
Type of Last Military
Discharge
Prior Felony
Incarcerations
County in Which
Sentence was
Imposed
Prior Jail Time
12
Prior Prison Time
40
41
42
20
Sentence
1
Abt Associates Inc.
Definition
A unique number that identifies an offender within the agency for this
admission and all subsequent admissions.
The offender’s unique, fingerprint-supported state identification number
The unique identification number given by the Federal Bureau of
Investigation/ Interstate Identification Index to each offender
The offender’s first name
The offender’s last name
9-digit social security number
The offender’s date of birth
The offender’s biological sex
The offender’s race
Is the offender of Hispanic origin?
The highest academic grade level the offender completed prior to
admission to prison on the current sentence
Did the offender ever served in the U.S. Armed Forces?
The date the offender was discharged from the U.S. Armed Forces for the
final time
The type of discharge the inmate received from the U.S. Armed Forces
Core
Variable
Was the offender ever sentenced to confinement for a felony as a juvenile
or adult prior to his/her current prison admission?
The county where the court imposing the current sentence is located
The length of time served in jail prior to the date of admission (Variable 8)
and credited to prison service for the current sentence
The length of time served in prison prior to the date of admission (Variable
8) and credited to prison service for the current sentence
2015 NCRP Data Request Instructions ▌pg. 10
Category
#
13
Name
Offenses
14a
Offense with
Longest Maximum
Sentence
Sentence Length for
Variable 14a
Offense
Total Maximum
Sentence Length
Indeterminate
Sentence
Determinant
Sentence
Mandatory Minimum
Sentence
Truth in Sentencing
Law Restriction
14b
15
31a
31b
31c
31d
32
Admission
to Prison
8
9
10
Abt Associates Inc.
Length of CourtImposed Sentence
to Community
Supervision
Date of Admission to
Prison
Type of Admission
to Prison
Jurisdiction on Date
of Admission
Definition
Crime(s) for which the offender was admitted to prison on the current
sentence(s), including the number of counts for each offense.
Of the crimes coded in Variable 13, the ONE crime for which the inmate
received the longest sentence
The maximum sentence as stated by the court that the offender is required
to serve for the offense listed in Variable 14a
Core
Variable
The longest length of time as stated by the court that the offender could be
required to serve for all offenses specified in Variable 13 (Offenses)
Does the total maximum sentence (Variable 15) include an indeterminate
sentence?
Does the total maximum sentence (Variable 15) include a determinate
sentence?
Does the total maximum sentence (Variable 15) include a mandatory
minimum sentence?
Is the total maximum sentence (Variable 15) restricted by a Truth in
Sentencing Law mandating that a certain percentage of the court- imposed
sentence be served in prison?
The amount of time which the court states that the offender is required to
serve under community supervision after release from prison
The most recent date the offender was admitted into the custody of the
state prison system on the current sentence
The reason an offender entered into the physical custody of a correctional
facility on the date provided in Variable 8 (Admission Date) of the current
record
The state with the legal authority to enforce the prison sentence
2015 NCRP Data Request Instructions ▌pg. 11
Category
#
17
Additional
Sentences
Since
Admission
18
Release
from prison
23a
19
25
21
22a
22b
23b
24
Abt Associates Inc.
Name
Location where
Offender is to Serve
Sentence
Additional Offenses
Since Admission
Date
Additional Sentence
Time Since
Admission
Date of Release
from Prison
Type of Release
from Prison
AWOL or Escape
Community Release
Prior to Prison
Release
Number of Days on
Community Release
Location at Time of
Prison Release
Agencies Assuming
Custody at Time of
Prison Release
Definition
The type of facility in which the offender will be incarcerated to serve time
for his/her crime.
Core
Variable
Any additional offense imposed after the date of admission (Variable 8),
regardless of the date of the crime.
The maximum time the inmate may be incarcerated consecutive to the
sentence length coded in Variable 15
The most recent calendar date that the state's prison custody terminated
The method of, or reason for, departure from the custody of your prison
system on the reported date of release
Was the offender AWOL or did (s)he escape while serving sentences?
Prior to release from the custody of a prison system, was the offender
concurrently under community based supervision or placement?
The number of days the inmate was on community release prior to release
from prison (if Variable 22a is yes)
The type of facility that had been used for the custody or care of the
offender just prior to release
The type and location of agency that assumes custody (physical or
supervisory) over an inmate's freedom at the time of prison release
2015 NCRP Data Request Instructions ▌pg. 12
Part D (Prison Custody) Instructions
The data file you produce for Part D should contain one data record for each sentenced inmate
under physical custody, regardless of sentence length or jurisdiction, on December 31, 2015.
Include in Part D:
Sentenced inmates in your prison facilities5, regardless of jurisdiction or sentence length.
Sentenced inmates under your jurisdiction held in county or local jails.
Sentenced inmates under your jurisdiction held in in-state or out-of-state private prisons,
including both privately owned facilities and facilities operated by a private entity under contract
to the state.
Any inmate in the above categories who was temporarily released (less than 30 days) from a
facility.
Exclude from Part D:
Sentenced inmates under your jurisdiction held in Federal facilities or another state’s facilities.
Unsentenced inmates held in your prison facilities (e.g., civil commitments, inmates awaiting
trial).
Inmates who have escaped and are not in custody.
The variables requested in the Part D data set are listed on the next page. Most of these variables are
also in the Part A and B requests. Refer to the Appendix for additional information on these variables.
5
Prison facilities include prisons, penitentiaries, and correctional institutions; boot camps; prison farms;
reception, diagnostic, and classification centers; release centers, halfway houses, and road camps; forestry
and conservation camps; vocational training facilities; prison hospitals; and drug and alcohol treatment
facilities for prisoners. For inmates under home confinement, a private residence is not considered a prison
facility.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 13
The Part D (Prison Custody) variables are listed below in the table. If you have limited resources for responding to this data request, please
focus on the core variables. Additional information on the variables is in the Appendix.
Category
Offender
#
2
Name
Inmate ID Number
30
39
State ID Number
FBI Number
36
37
46
3
4
5
6
7
11
First Name
Last Name
SSN
Date of Birth
Sex
Race
Hispanic Origin
Highest Grade
Completed
Prior Military Service
Date of Last Military
Discharge
Type of Last Military
Discharge
Prior Felony
Incarcerations
County in Which
Sentence was
Imposed
Prior Jail Time
12
Prior Prison Time
40
41
42
20
Sentence
1
Abt Associates Inc.
Definition
A unique number that identifies an offender within the agency for this
admission and all subsequent admissions.
The offender’s unique, fingerprint-supported state identification number
The unique identification number given by the Federal Bureau of
Investigation/ Interstate Identification Index to each offender
The offender’s first name
The offender’s last name
9-digit social security number
The offender’s date of birth
The offender’s biological sex
The offender’s race
Is the offender of Hispanic origin?
The highest academic grade level the offender completed prior to
admission to prison on the current sentence
Did the offender ever served in the U.S. Armed Forces?
The date the offender was discharged from the U.S. Armed Forces for the
final time
The type of discharge the inmate received from the U.S. Armed Forces
Core
Variable
Was the offender ever sentenced to confinement for a felony as a juvenile
or adult prior to his/her current prison admission?
The county where the court imposing the current sentence is located
The length of time served in jail prior to the date of admission (Variable 8)
and credited to prison service for the current sentence
The length of time served in prison prior to the date of admission (Variable
8) and credited to prison service for the current sentence
2015 NCRP Data Request Instructions ▌pg. 14
Category
#
13
Name
Offenses
14a
Offense with
Longest Maximum
Sentence
Sentence Length for
Variable 14a
Offense
Total Maximum
Sentence Length
Indeterminate
Sentence
Determinant
Sentence
Mandatory Minimum
Sentence
Truth in Sentencing
Law Restriction
14b
15
31a
31b
31c
31d
32
Prison
Admission
8
9
10
51
Abt Associates Inc.
Length of CourtImposed Sentence
to Community
Supervision
Date of Admission to
Prison
Type of Admission
to Prison
Jurisdiction on Date
of Admission
Custodial Security
Level
Definition
Crime(s) for which the offender was admitted to prison on the current
sentence(s), including the number of counts for each offense.
Of the crimes coded in Variable 13, the ONE crime for which the inmate
received the longest sentence
The maximum sentence as stated by the court that the offender is required
to serve for the offense listed in Variable 14a
Core
Variable
The longest length of time as stated by the court that the offender could be
required to serve for all offenses specified in Variable 13 (Offenses)
Does the total maximum sentence (Variable 15) include an indeterminate
sentence?
Does the total maximum sentence (Variable 15) include a determinate
sentence?
Does the total maximum sentence (Variable 15) include a mandatory
minimum sentence?
Is the total maximum sentence (Variable 15) restricted by a Truth in
Sentencing Law mandating that a certain percentage of the court- imposed
sentence be served in prison?
The amount of time which the court states that the offender is required to
serve under community supervision after release from prison
The most recent date the offender was admitted into the custody of the
state prison system on the current sentence
The reason an offender entered into the physical custody of a correctional
facility on the date provided in Variable 8 (Admission Date) of the current
record
The state with the legal authority to enforce the prison sentence
Level of security at which the offender is held in prison
2015 NCRP Data Request Instructions ▌pg. 15
Category
#
17
Anticipated
Release
from Prison
33
34
35
Facility
Additional
Sentences
Since
Admission
38
18
19
Abt Associates Inc.
Name
Location where
Offender is to Serve
Sentence
Parole
Hearing/Eligibility
Date
Projected Release
Date
Mandatory Release
Date
Facility Name
Additional Offenses
Since Admission
Date
Additional Sentence
Time Since
Admission
Definition
The type of facility in which the offender will be incarcerated to serve time
for his/her crime.
Core
Variable
The date the offender is eligible for review by an administrative agency
such as a parole board, to determine whether he or she will be released
from prison
The projected date on which the offender will be released from prison
The date the offender by law must be conditionally released from prison
Name of the facility holding the offender at year-end
Any additional offense imposed after the date of admission (Variable 8),
regardless of the date of the crime.
The maximum time the inmate may be incarcerated consecutive to the
sentence length coded in Variable 15.
2015 NCRP Data Request Instructions ▌pg. 16
Part E (Post Confinement Community Supervision Admissions)
Instructions
The data file you produce for Part E should contain one data record for each admission of an
offender to a term of post-confinement community supervision (PCCS) to your state during
2015. PCCS means sentenced offenders serving a period of community supervision immediately after
release from prison. Only include admissions to PCCS of offenders under the legal authority of your
state; do not include interstate compact cases in which only supervisory responsibility is transferred to
your state but legal authority is retained by another state.
Include in Part E:
Admissions to community supervision for the purpose of completing a prison term in the
community. Most states refer to this as parole; your state may use other terminology. Examples
include:
An offender is released from a prison facility by the decision of a parole board or other
authority to the caseload of a community supervision authority (e.g., parole agency,
probation agency, corrections department). Most states call this a discretionary prison
release.
An offender has a mandatory release from prison to the caseload of a community
supervision authority (e.g., parole agency, probation agency, corrections department).
Admissions to community supervision resulting from a community supervision sentence that
begins immediately upon release from prison. This includes what some states refer to as a split
sentence or shock probation. Examples include:
An offender begins serving a court-imposed sentence of community supervision
following release from prison.
Re-admissions to community supervision following a revocation from community supervision
and a subsequent release from prison to complete the sentence in the community.
Admissions of offenders to community supervision in your state following a term of confinement
in another state when that state transfers legal authority of the offender to your state.
Exclude from Part E:
Admissions to community supervision that are not immediately preceded by a term of
confinement.
Admissions to prison facilities.6
6
Prison facilities include prisons, penitentiaries, and correctional institutions; boot camps; prison farms;
reception, diagnostic, and classification centers; release centers, halfway houses, and road camps; forestry
and conservation camps; vocational training facilities; prison hospitals; and drug and alcohol treatment
facilities for prisoners.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 17
Inmates re-entering parole or supervised release after a leave that was NOT a revocation.
Example: An offender serving a term of supervision is picked up on a technical violation
and sent back to prison for a “shock” term. The offender is never released from
supervision and the supervising agency has jurisdiction over the offender the entire time.
Interstate compact cases where only supervisory responsibility is transferred to your state but
legal jurisdiction is retained by another state.
The variables requested in the Part E data set are listed below. Refer to the Appendix for additional
information on these variables.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 18
The Part E (Post-Confinement Community Supervision Admissions) variables are listed below in the table. If you have limited resources for
responding to this data request, please focus on the core variables. Additional information on the variables is in the Appendix.
Category
Offender
#
2
Name
Inmate ID Number
30
39
State ID Number
FBI Number
36
37
46
47
First Name
Last Name
SSN
Residential Street
Address
Residential City
Residential State
Residential Zip
Code
Date of Birth
Sex
Race
Hispanic Origin
Highest Grade
Completed
Prior Military Service
Date of Last Military
Discharge
Type of Last Military
Discharge
County in Which
Sentence was
Imposed
48
49
50
3
4
5
6
7
40
41
42
Sentence
1
Abt Associates Inc.
Definition
A unique number that identifies an offender within the agency for this
admission and all subsequent admissions.
The offender’s unique, fingerprint-supported state identification number
The unique identification number given by the Federal Bureau of
Investigation/ Interstate Identification Index to each offender
The offender’s first name
The offender’s last name
9-digit social security number
Street address of residential address prior to imprisonment
Core
Variable
City of residence prior to imprisonment
State of residence prior to imprisonment
Zip code of residence prior to imprisonment
The offender’s date of birth
The offender’s biological sex
The offender’s race
Is the offender of Hispanic origin?
The highest academic grade level the offender completed prior to
admission to prison on the current sentence
Did the offender ever served in the U.S. Armed Forces?
The date the offender was discharged from the U.S. Armed Forces for the
final time
The type of discharge the inmate received from the U.S. Armed Forces
The county where the court imposing the current sentence is located
2015 NCRP Data Request Instructions ▌pg. 19
Category
Release
from Prison
#
13
Name
Offenses
23a
Date of Release
from Prison
Type of Release
from Prison
Agencies Assuming
Custody at Time of
Prison Release
Date of Admission to
Post-Confinement
Community
Supervision
Type of Admission
to Post-Confinement
Community
Supervision
25
24
Admission
to PCCS
43
44
Abt Associates Inc.
Definition
Crime(s) for which the offender was admitted to prison on the current
sentence(s)
The most recent calendar date that the state's prison custody terminated
The method of, or reason for, departure from the custody of your prison
system on the reported date of release
The type and location of the agency that assumes custody (physical or
supervisory) over an inmate's freedom at the time of prison release
Core
Variable
The date an offender entered into post-confinement community supervision
The reason an offender entered into post-confinement community
supervision on the date provided in Variable 43 (Date of Admission to PostConfinement Community Supervision) of the current record
2015 NCRP Data Request Instructions ▌pg. 20
Part F (Post-Confinement Community Supervision Releases)
Instructions
The data file you produce for Part F should contain one data record for each release of an offender
serving a term of post-confinement community supervision (PCCS) during 2015. PCCS means
sentenced offenders serving a period of community supervision immediately after release from
prison. Only include releases from PCCS of offenders under the legal jurisdiction of your state; do not
include interstate compact cases in which your state only had supervisory responsibility and another
state retained legal jurisdiction over the offender.
NCRP defines PCCS releases as including:
Discharges
Returns to prison or jail resulting from a revocation, pending revocation, or a new sentence
Transfer of legal authority over an offender from your state to another state
Deaths
Include in Part F:
Releases from community supervision when the offender was completing his prison sentence.
Examples include:
An offender is returned to prison while on parole, supervised release, mandatory
supervised release, or other types of post-confinement community supervision.
An offender is discharged after completing parole, supervised release, mandatory
supervised release, or other types of conditional release.
An offender is discharged after completing parole, supervised release, mandatory
supervised release, or other types of conditional release, but then begins serving a courtimposed sentence of community supervision.
Releases from community supervision that resulted from a separate sentence that began following
release from prison. Examples include:
An offender completes a court-imposed term of probation after serving a term of
incarceration.
An offender is returned to prison while serving a court-imposed term of probation after
serving a prison term.
Transfer of legal authority from your state to another state of an offender on community
supervision following a prison term.
Exclude from Part F:
Releases from community supervision when the offender did not serve a term of incarceration
immediately preceding the term of community supervision.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 21
Releases from prison facilities.7
Temporary revocations where the inmate is not removed from supervision, and not re-admitted
into a facility.
Example: An offender serving a term of supervision is picked up on a technical violation
and sent back to prison for a “shock” term. The offender is never released from
supervision and the supervising agency has jurisdiction over the offender the entire time.
Releases of un-sentenced inmates who are being supervised in the community but who have not
served a sentenced term of incarceration.
Interstate compact cases in which your state only had supervisory responsibility and another state
retained legal jurisdiction over the offender.
The variables requested in the Part F data set are listed below. Refer to the Appendix for additional
information on these variables.
7
Prison facilities include prisons, penitentiaries, and correctional institutions; boot camps; prison farms;
reception, diagnostic, and classification centers; release centers, halfway houses, and road camps; forestry
and conservation camps; vocational training facilities; prison hospitals; and drug and alcohol treatment
facilities for prisoners.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 22
The Part F (Post-Confinement Community Supervision Releases) variables are listed below in the table. If you have limited resources for
responding to this data request, please focus on the core variables. Additional information on the variables is in the Appendix.
Category
Offender
#
2
Name
Inmate ID Number
30
39
State ID Number
FBI Number
36
37
46
3
4
5
6
7
First Name
Last Name
SSN
Date of Birth
Sex
Race
Hispanic Origin
Highest Grade
Completed
Prior Military Service
Date of Last Military
Discharge
Type of Last Military
Discharge
County in Which
Sentence was
Imposed
Offenses
40
41
42
Sentence
1
13
Release
from Prison
23a
25
Abt Associates Inc.
Date of Release
from Prison
Type of Release
from Prison
Definition
A unique number that identifies an offender within the agency for this
admission and all subsequent admissions.
The offender’s unique, fingerprint-supported state identification number
The unique identification number given by the Federal Bureau of
Investigation/ Interstate Identification Index to each offender
The offender’s first name
The offender’s last name
9-digit social security number
The offender’s date of birth
The offender’s biological sex
The offender’s race
Is the offender of Hispanic origin?
The highest academic grade level the offender completed prior to
admission to prison on the current sentence
Did the offender ever served in the U.S. Armed Forces?
The date the offender was discharged from the U.S. Armed Forces for the
final time
The type of discharge the inmate received from the U.S. Armed Forces
Core
Variable
The county where the court imposing the current sentence is located
Crime(s) for which the offender was admitted to prison on the current
sentence(s)
The most recent calendar date that the state's prison custody terminated.
The method of, or reason for, departure from the custody of your prison
system on the reported date of release
2015 NCRP Data Request Instructions ▌pg. 23
Category
#
24
Admission
to PCCS
43
44
Release
from PCCS
26
27
28
45
Abt Associates Inc.
Name
Agencies Assuming
Custody at Time of
Prison Release
Date of Admission to
Post-Confinement
Community
Supervision
Type of Admission
to Post-Confinement
Community
Supervision
Date of Release
from PostConfinement
Community
Supervision
Type of Release
from PostConfinement
Community
Supervision
Supervision Status
Just Prior to
Release
County Where
Offender was
Released / County
Where PCCS Office
is Located
Definition
The type and location of the agency that assumes custody (physical or
supervisory) over an inmate's freedom at the time of prison release
Core
Variable
The date an offender entered into post-confinement community supervision.
The reason an offender entered into post-confinement community
supervision on the date provided in Variable 43 (Date of Admission to PostConfinement Community Supervision) of the current record
The date of discharge or termination from post-confinement community
supervision jurisdiction for any reason, including returning the offender to
prison
The reason for the termination of post-confinement community supervision
jurisdiction that occurred on the date provided in Variable 26
The level of contact the PCCS agency had with the offender during the year
prior to release from PCCS
The county where the offender was released from post-confinement
community supervision on the date in Variable 26. If not available, report
the county where the PCCS office to which the offender reported before exit
is located.
2015 NCRP Data Request Instructions ▌pg. 24
Appendix.
Additional Information on NCRP Variables
Variable 1: County in Which Sentence Was Imposed
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
The county where the court imposing the current sentence is located. If there are multiple
counties of commitment, use the one which corresponds with the offense for which the person
received the longest maximum sentence.
Codes / Coding Information
If possible, use either the name of the county or the 5-digit county FIPS code (available at
http://www.itl.nist.gov/fipspubs/co-codes/states.txt).
Variable 2: Inmate ID Number
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
A unique number that identifies an offender within the state department of corrections.
Parole or other community supervision agencies that do not have access to the department of
corrections inmate identification number can provide their own agency’s unique identification
number for the offender.
Additional Information
Do not use sequence numbers for identification numbers unless you can identify each inmate by
the sequence number and use the same sequence number for the inmate's every movement into or
out of the corrections system.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 25
All information that can identify individuals will be held strictly confidential by Abt Associates
and the Bureau of Justice Statistics, per the requirements of Title 42, United States Code, Sections
3735 and 3789g.
Variable 3: Date of Birth
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
The offender’s date of birth
Report partial dates if the day or month is not known.
Variable 4: Sex
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
The offender’s biological sex
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
(2)
(9)
Male
Female
Not known
Variable 5: Race
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 26
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
The offender’s race
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(9)
White. A person having origins in any of the original people of Europe, North Africa, or
the Middle East.
Black. A person having origins in any of the black racial groups of Africa.
American Indian / Alaskan Native. A person having origins in any of the original people
of North America and South America (including Central America), and who maintains
tribal affiliations or community attachment.
Asian. A person having origins in any of the original peoples of the Far East, Southeast
Asia, or the Indian subcontinent including, for example, Cambodia, China, India, Japan,
Korea, Malaysia, Pakistan, the Philippine Islands, Thailand, and Vietnam.
Native Hawaiian / Pacific Islander. A person having origins in any of the original
peoples of Hawaii, Guam, Samoa, or other Pacific Islands.
Other categories in your information system. Other single-race categories not listed
above which are in your information system.
Two or more races. A person who identifies with more than one racial category and/or a
person who identifies as multi-racial.
Not known. Racial category is not known.
Additional Information
Hispanic origin is a cultural characteristic rather than racial characteristic (see Variable 6).
Persons of Hispanic origin can be black, white or some other racial group. When the information
is available, please code the racial characteristic of persons of Hispanic origin.
If the inmate’s race can be determined but does not fit one of the above categories, then code as
“other categories in your information system.”
Variable 6: Hispanic Origin
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 27
Definition
Whether the offender is of Hispanic origin
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
(2)
(9)
Hispanic or Latino origin. A person of Mexican, Puerto Rican, Cuban, Central American,
South American, or other Spanish culture or origin, regardless of race.
Not of Hispanic origin.
Not known (Hispanic origin is not known).
Variable 7: Highest Grade Completed
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
The highest academic grade level completed by the offender before being admitted to prison on
the current sentence.
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(99)
8th Grade or Less (level of education did not exceed 8th grade, including having never
attended school).
Some High School (grade unspecified or grade completed is not available but it is known
that the inmate entered high school or started 9th grade).
9th Grade
10th Grade
11th Grade
12th Grade or GED
Some College (any person who attended college but did not graduate).
College Degree (any person who completed college or had some post-graduate
education).
Special/Ungraded (including Special education, vocational education/rehabilitation,
occupational education/rehabilitation, academic in an ungraded system, technical
training, or education in an ungraded system).
Not known (level of education is not known).
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 28
Additional Information
Do not report any educational work completed during incarceration on the current sentence.
Do not report competency level.
Variable 8: Date of Admission to Prison
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Definition
The most recent date the inmate was admitted into the custody of the state prison system on the
current sentence.
Report partial dates if the day or month is not known.
Additional Information
Do not provide the sentencing date as the date of admission unless correctional custody began
immediately after sentencing. Admission date should never be prior to the sentencing date.
Offenders exiting from post confinement community supervision and returning to prison as
violators should be included in both the Part A (prison admission) and Part F (post confinement
community supervision release) files.
Prisoner admission data should be provided for sentenced state prisoners housed in local jails.
The date of admission for prison inmates housed in local jails is the date on which the prison
system assumed jurisdiction, often the date of sentencing. Once you submit an admission record
to NCRP for a sentenced state prisoner who is housed in a local jail, do not later report his/her
transfer from jail to prison as an admission.
Examples
A person held in a local jail is sentenced on April 3, 2009. Due to prison overcrowding, he begins
serving his sentence in the local jail immediately after sentencing. The date of admission to prison
is reported as April 3, 2009.
A prisoner held in a local jail is sentenced on April 3, 2009. Due to prison overcrowding, she
begins serving her sentence in a local jail immediately after sentencing. She is transferred and
physically enters prison on October 28, 2009. No record of any kind is created for the October
transfer. Instead, a Part A record is created with April 3, 2009 as the date of admission.
A person was admitted originally on June 11, 2003. He was released to parole supervision in
2005 and readmitted to prison August 7, 2009 as the result of a parole revocation. For the Part A
(prison admission) record, the date of admission is August 7, 2009.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 29
Variable 9: Type of Admission to Prison
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Definition
The reason an offender entered into the physical custody of a correctional facility on the date
provided in Variable 8 of the current record.
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(10)
Court Commitment. A person being admitted to prison on one or more new sentences; the
person is being confined for the first time on this/these particular sentence(s) and is not
being re-admitted on any previous sentences still in effect.
(20)
Returned from Appeal or Bond. An offender's re-entry into prison after an absence on
appeal bond during which his/her sentence time was not running. Do not create a new
admission record upon an inmate's return if the inmate's sentence time continued to run
while he/she was on appeal bond.
(30)
Transfer. The admission of a person from the custody of another detaining authority to
continue serving the same sentence.
(46)
Discretionary Release Revocation, New Sentence. Discretionary release occurs when an
inmate is conditionally released by the decision of a parole board or other authority.
Revocation is the administrative action of a supervising agency removing a person from
supervision status in response to a violation of conditions of supervision. If discretionary
release is revoked because of a new sentence, use code 46.
Discretionary Release Revocation, No New Sentence. Discretionary release occurs when
an inmate is conditionally released by the decision of a parole board or other authority.
Revocation is the administrative action of a supervising agency removing a person from
supervision status in response to a violation of conditions of supervision. If discretionary
release is revoked because of a technical violation, use code 47.
Discretionary Release Revocation, No Information. Discretionary release occurs when an
inmate is conditionally released by the decision of a parole board or other authority.
Revocation is the administrative action of a supervising agency removing a person from
supervision status in response to a violation of conditions of supervision. If discretionary
release has been revoked and the reason is not known, use code 49.
(47)
(49)
(56)
Mandatory Conditional Release Revocation, New Sentence. Mandatory conditional
release occurs when an inmate must, by law, be conditionally released from prison to
serve the remainder of their sentence in the community. Revocation is the administrative
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 30
(57)
(59)
action of a supervising agency removing a person from supervision status in response to a
violation of conditions of supervision. This type of release may also be called "mandatory
parole" or "supervised mandatory release." Use code 56 if mandatory conditional release
is revoked because of a new sentence.
Mandatory Conditional Release Revocation, No New Sentence. Mandatory conditional
release occurs when an inmate must, by law, be conditionally released from prison to
serve the remainder of their sentence in the community. Revocation is the administrative
action of a supervising agency removing a person from supervision status in response to a
violation of conditions of supervision. This type of release may also be called "mandatory
parole" or "supervised mandatory release." Use code 57 if mandatory conditional is
revoked because of a technical violation.
Mandatory Conditional Release Revocation, No Information. Mandatory conditional
release occurs when an inmate must, by law, be conditionally released from prison to
serve the remainder of their sentence in the community. Revocation is the administrative
action of a supervising agency removing a person from supervision status in response to a
violation of conditions of supervision. This type of release may also be called "mandatory
parole" or "supervised mandatory release. Use code 59 if mandatory conditional release is
revoked and the reason is not known.
(65)
Court Commitment/Suspended Sentence Imposed. Use this code if the admission is the
result of the court's imposition of a previously suspended sentence.
(66)
Escapee/AWOL Returned, New Sentence. Use this code if an escaped inmate is returned
with a new sentence. The new sentence may be for escaping or another offense.
Escapee/AWOL Returned, No New Sentence. Use this code if an escaped inmate is
returned and it is not known if there is a new sentence.
Escapee/AWOL Returned, No Information. Use this code if an escaped inmate is returned
and it is not known if there is a new sentence.
(67)
(69)
(70)
(80)
(90)
(86)
(87)
Court Commitment/Discretionary Release Status, Pending Revocation. Use this code if
the inmate has violated the conditions of discretionary release supervision but his/her
discretionary release has not been formally revoked.
Court Commitment/Mandatory Conditional Release Status, Pending Revocation. Use this
code if the inmate has violated the conditions of mandatory conditional release
supervision, but his/her conditional release has not been formally revoked.
Court Commitment/Probation Status, Pending Revocation. Use this code if the inmate
has violated the conditions of probation, but his/her probation has not been formally
revoked.
Probation Revocation, New Sentence. Probation Revocation is a court order taking away
a person's probationary status in response to a violation of conditions of probation. Use
this code if the probation was revoked as a result of a new sentence.
Probation Revocation, No New Sentence. Probation Revocation is a court order taking
away a person's probationary status in response to a violation of conditions of probation.
Use this code if probation is revoked due to a technical violation.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 31
(89)
Probation Revocation, No Information. Probation Revocation is a court order taking
away a person's probationary status in response to a violation of conditions of probation.
Use this code if the probation was revoked and the reason is not known.
(88)
Other. If a unique code cannot be assigned, use code 88 and document the types of
admission included in this category.
(92)
Unsentenced Commitment.
(99)
Not Known. Use this code if the type of admission is Not Known.
Additional Information
For Code 10 (Court Commitment):
Include as a court commitment inmates sentenced to prison for brief periods of time,
usually 90-180 days, after which they are either released to probation or remain in prison.
If, at the end of the "shock" period, the court commits the offender to prison to continue
serving sentence, do not report him/her again as an admission.
Exclude from the court commitment category: all revocations of probation, parole or
other conditional release with or without a new sentence for a new offense; all transfers
unless the inmate has completed all previous sentences and is beginning to serve time on
a new sentence; and all returns from escape or unauthorized departures.
For Code 20 (Returned from Appeal or Bond):
Do not create a new admission record upon an inmate's return if the inmate's sentence
time continued to run while he/she was on appeal bond.
For Code 30 (Transfer):
Include inmates admitted from a long term stay in a hospital, mental health facility or
another state or federal prison.
Do not provide records for movements from prison facility to prison facility within your
state.
Do not report the return of an inmate sent temporarily to another state to stand trial.
Do not include inmates who have completed a sentence in another state and are
transferred to your state to begin serving a different sentence. Code them as court
commitments, post-confinement community release revocations or other, as appropriate.
Codes 46, 47, and 49 (Discretionary Release Revocation) are limited to those cases where
revocation proceedings have been completed.
Codes 56, 57, and 59 (Mandatory Conditional Release) also are only applicable to those cases
where revocation proceedings have been completed.
Examples
Court Commitment (Code 10)
A person is sentenced by the court for murder and transported to a state correctional
institution to begin serving her sentence. The correct code is "10" court commitment.
A person is sentenced by the court for murder and transported to a state correctional
facility to begin serving his/her sentence. This person is still on parole for a robbery he
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 32
committed four years ago but his parole revocation hearing has not been held yet. This
admission is not a court commitment. Use code 70 or 80 to report admission type for this
inmate.
A person is sentenced in 2001 to serve three years for burglary. She is conditionally
released after one year and completes her time on parole. She is now being incarcerated
for a burglary for which she has never served a sentence. The correct code is 10, "court
commitment."
An offender receives a sentence of five years, the first 120 days to be served in prison,
the remainder on probation. A Prison Admission record should be created and Variable 9
coded as 10, "court commitment."
Returned from Appeal or Bond (Code 20)
An inmate in prison is granted an appeal and released on bond. His sentence time is not
running. His guilt and sentence are later reaffirmed and he returns to prison to resume
serving his sentence. The admission type is code 20, "return from appeal bond."
Transfer (Code 30)
An inmate serving a prison sentence was declared insane and surrendered to the custody
of the State Department of Mental Health. This movement constituted a transfer release.
This year the inmate is found sane and returns to prison to resume serving the sentence. A
Prison Admission record should be created and the type of admission coded as 30,
"Transfer."
An inmate is sentenced in California to serve 5 years for burglary and enters a California
prison to begin serving her sentence. During the report year, she is transferred to a
Nevada prison for protective custody. This movement is a prison release type, "Transfer"
for California. Nevada would report this inmate's admission as code 30, "Transfer."
An inmate serving a prison sentence in Rhode Island is temporarily released to Vermont
to stand trial for charges in that state. The inmate is found guilty and returned one week
later to Rhode Island to continue serving his/her time. No admission or release record is
created by either state.
A Rhode Island inmate is serving a two-year sentence. After serving one year of his
sentence, he is sent to Vermont to serve the balance of his sentence. The correct response
for each state is as follows:
Rhode Island creates a prison release record - Variable 25 (type of prison release)
is coded as 15, "Transfer."
Vermont creates a prison admission record - Variable 9 is coded 30, "Transfer."
In February of the report year, an inmate is admitted to a Maryland State prison to begin
serving a three year sentence for armed robbery. In June of the same year, he is
transferred to a county detention facility for safekeeping. An admission record is created
when the inmate is admitted in February. No admission or release record is created when
the inmate is transferred to the county facility because he is still serving the state sentence
at the county facility and he is still in the state of Maryland.
A Maine inmate is transferred during the report year from the Maine Correctional Center
(a state facility) to the Maine State Prison. The correct response is to create no admission
or release record for inmates that are transferred among state facilities within your state.
Discretionary Release Revocations (Codes 46, 47, 49)
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 33
While on discretionary release, the offender commits an armed robbery and is sentenced
to serve time for that offense. His discretionary release is revoked, and he enters prison to
begin serving time on the new sentence. Code 46, "discretionary release revocation, new
sentence" is the correct code.
Mandatory Conditional Release Revocations(Codes 56, 57, 59):
While on mandatory conditional release, an offender fails to report to his/her supervising
authority. Her conditional release is revoked and she returns to prison to continue serving
time on the original sentence. Code 57, "mandatory conditional release revocation, no
new sentence" is the correct code to use in this instance.
Escape/AWOL Return (Codes 66, 67, 69):
An inmate escaped from prison in December, last year. A release record was created for
that calendar year. He was located and returned to prison in June this year with no new
sentence. An admission record is created and the admission type is coded 67, "escapee
returned, no new sentence."
An inmate escaped from prison in June. While on escape status, he commits a burglary
and is arrested and placed in jail. He is found guilty of burglary, sentenced, and returned
to prison in December. His admission type is code 66, "escapee returned, new sentence."
Court Commitment/Discretionary Release Status, Pending Revocation (Code 70)
An offender violates the conditions of his discretionary release and is accused of
committing a new offense. He is returned to prison. The new charges are pending. The
discretionary release revocation hearing has not been held yet. The correct code is 70,
"discretionary release status, pending revocation."
Variable 10: Jurisdiction on Date of Admission
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Definition:
The state with the legal authority to enforce the prison sentence on the date of admission in
Variable 8.
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
State FIPS Codes (available at http://www.itl.nist.gov/fipspubs/fip5-2.htm)
(52)
Jurisdiction is shared between states
(57)
Federal Prison System has jurisdiction
(60)
State not known
(99)
Not known
Examples
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 34
An inmate is convicted of murder in Maryland and sentenced to a 10-year prison term. He begins
serving his sentence in a Virginia prison to ensure protective custody. Maryland is the correct
value.
Variable 11: Prior Jail Time
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Definition
The length of time served in jail prior to the date of admission (provided in Variable 8) and
credited to prison service for the current sentence.
Additional Information
If it is known that some prior time had been served but prior jail time cannot be distinguished
from prior prison time, include all prior time in the prior prison time category (see Variable 12).
Examples
A man was arrested and charged with burglary on January 1 of this year. He spent two months in
jail awaiting trial. He was convicted on March 1 and was sentenced to serve two years in prison.
The judge allows his time in jail to be credited toward his total sentence. The correct value for
Variable 11 is two months.
A man was arrested and charged with burglary on January 1 of this year. He spent two months in
jail awaiting trial. He was convicted and sentenced on March 1. The judge states that his prison
time begins running as of his date of sentencing. The correct code for Variable 11 is zero days,
because no time in jail was credited toward his sentence.
On July 1, 2005 an inmate was admitted to a local jail, due to overcrowding, to begin serving a
5-year sentence for drug trafficking. He was released to post-confinement community supervision
(PCCS) on December 15, 2006. He is now being admitted to prison on a PCCS revocation and
must serve the remainder of his drug trafficking sentence in prison. The time he served in jail for
this offense, prior to his release to PCCS, counts toward his total time incarcerated on the current
sentence and must be reported. The correct value to report is one year, five months, and 15 days.
Variable 12: Prior Prison Time
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 35
Definition
The length of time served in prison prior to the date of admission (provided in Variable 8) and
credited to prison service for the current sentence.
Additional Information
If it is known that some prior time had been served but prior jail time cannot be distinguished
from prior prison time, include all prior time in the prior prison time category.
Only time spent in confinement and credited against the current sentence should be reported.
Examples
A man is admitted to prison on June 1, 2003 to begin serving a 10-year term for armed robbery.
He is paroled July 10, 2010. He violates the conditions of his parole and returns to prison this
year to complete his sentence. The time he served in prison prior to his parole counts toward his
total time served for this offense and must be reported. The correct value to report is 7 years, 1
month, and 10 days.
A man is admitted to prison on June 1, 2003 to begin serving a 10-year term for armed robbery.
His sentence is commuted on July 10, 2010 and he is unconditionally released. However, he
commits a new offense this year and is sentenced to serve 3 years in prison. His previous sentence
does not affect this new sentence in any way. The correct value to report is 0 days.
Variable 13: Offenses
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
Crime(s) for which the offender was admitted to prison on the current sentence(s).
Include the number of counts of each offense.
Codes / Coding Information
Use your state's own offense codes. NCRP staff will re-code your state’s offense codes into the
NCRP offense codes (available at https://www.ncrp.info/SitePages/FAQs.aspx).
Additional Information
Please submit offense code documentation along with data submission. This documentation
should include all of your states' offense codes and a description of each offense.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 36
For persons readmitted to prison, the original crime(s) in addition to any new crime(s) resulting in
the current sentence(s) should be indicated.
Variable 14a: Offense with Longest Maximum Sentence
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Definition
Of the crimes reported in Variable 13, this is the ONE crime for which the inmate received the
longest sentence.
Additional Information
If the inmate received the same maximum sentence length for two different offenses, provide the
one your state would designate as the "controlling," "driving," or "most serious" offense.
Variable 14b: Sentence Length for Variable 14a Offense
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Definition
The maximum sentence as stated by the court, that the offender is required to serve for the
offense listed in Variable 14a.
Codes / Coding Information
Report a life or a death sentence using either your agency’s codes or the following NCRP codes.
(99996)
(99997)
(99994)
(99993)
Maximum sentence is Life.
Maximum sentence is Death.
Maximum sentence is Life plus additional years.
Maximum sentence is Life without discretionary release.
Additional Information
This is the maximum sentence imposed by the court for one specific offense and should not
reflect any statutory or administrative sentence reductions.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 37
If the inmate has more than one sentence for the same type of offense, such as 2 years for one
burglary (or one count of burglary) and 3 years for another burglary (on another count of
burglary), the 3-year sentence would be reported for Variable 14b.
If the offense reported in Variable 14a is one for which the inmate was previously placed on postconfinement community supervision (e.g. parole or probation), provide the original maximum
sentence not the part of the sentence remaining to be served.
Please document any other code for life or death sentences that may appear on your file.
Examples
A man enters prison to begin serving time for three sentences. He received 5 years for burglary, 3
years for auto theft, and 1 year for a minor drug violation. The sentences are to be served
consecutively and result in a TOTAL maximum sentence of 9 years. However, for Variable 14a
and 14b, you need to indicate the one specific offense with the longest sentence. The correct
response for Variable 14a is your state code for burglary, and for 5 years for Variable 14b.
A man enters prison to begin serving time for two sentences. He received 5 years for burglary and
5 years for drug trafficking, both sentences to be served concurrently. In your state, burglary is
considered more serious and to be the "controlling" offense. Therefore, for Variable 14a, you
would provide your state code for burglary, and 5 years for Variable 14b.
A woman enters prison to begin serving time for three counts of burglary. She received 6 years
for the first count, 6 years for the second, and 4 years for the third, all to be served consecutively.
In Variable 14a, would be your state code for burglary, and 6 years for Variable 14b. Each count
is to be considered separately when it carries its own sentence length.
Variable 15: Total Maximum Sentence Length
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Definition
The longest length of time as stated by the court that the offender could be required to serve for
all offenses.
Codes / Coding Information
Report a life or a death sentence using either your agency’s codes or the following NCRP codes.
(99996)
(99997)
(99994)
(99993)
Maximum sentence is Life.
Maximum sentence is Death.
Maximum sentence is Life plus additional years.
Maximum sentence is Life without discretionary release.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 38
Additional Information
This is the maximum sentence imposed by the court and should not reflect any statutory or
administrative sentence reductions.
Do not subtract time credits or prior jail or prison time.
If all or a portion of a maximum sentence has been conditionally suspended (that is, the sentenced
person may in the future be required to serve the suspended sentence or only a portion under
certain circumstances), set the "Maximum Sentence" to the sum of the unsuspended and
suspended portions of the maximum sentence of each offense for which the inmate is currently in
prison.
Do not report unconditionally suspended sentences.
If all or a portion of a maximum sentence has been unconditionally suspended (that is, the person
cannot be required to serve the suspended sentence or any portion under any circumstances), use
as the "Maximum Sentence" only the unsuspended portions of the sentences.
For a split sentence or shock probation, set the maximum sentence to the sum of the prison and
probation segments of the sentence(s).
Provide the sum of sentences to be served consecutively. Do not add sentences to be served
concurrently.
Examples
An inmate receives a sentence of 3 years for possession of marijuana, 2 years conditionally
suspended. He will be released to post-confinement community supervision after being
imprisoned for one year. The correct value for Variable 15 is 3 years; that is, if his behavior is not
satisfactory, he will serve 3 years in prison.
A person receives a sentence of 5 years for burglary, one year unconditionally suspended. He will
receive no supervision during the one year regardless of his behavior. The correct value for
Variable 15 is 4 years.
A person receives a 10-year sentence for armed robbery, is paroled after 3 years, but returns to
prison on a technical violation 6 months later. The correct value for Variable 15 is 10 years,
reflecting his original maximum sentence.
A first offender receives a 5-year sentence for manslaughter, 90 days to be served in prison and
the remainder on probation. The correct value for Variable 15 is 5 years.
An offender enters prison to serve 6 years on a burglary conviction and 5 years on a drug
conviction. The two sentences are to be served consecutively. The correct value for Variable 15 is
11 years.
An offender enters prison to serve 6 years on a burglary conviction and 5 years on a drug
conviction. The two sentences are to be served concurrently. The correct value for Variable 15 is
6 years.
(There is no Variable 16)
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 39
Variable 17: Location Where Inmate is to Serve Sentence
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Definition
The type of facility in which the offender will be incarcerated to serve time for his crime.
The name of the facility can be provided instead. In this case, provide information in a separate
file that will enable Abt Associates to re-code the name of facility into the NCRP facility type
categories listed below.
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
(2)
(3)
(4)
(5)
(6)
(57)
(99)
State Prison Facility. A state administered confinement facility having custodial
authority over persons sentenced to confinement.
Local Jail. A confinement facility administered by an agency of the local government
intended for adults but sometimes also houses juveniles, which holds persons detained
pending adjudication and persons committed after adjudication usually with sentences of
a year or less.
Other Specify. All other facilities except those specified above which house sentenced
prisoners. Provide documentation for the type of facility included in this category.
Mental Hospital. A confinement facility for the diagnosis or treatment of mentally ill
patients.
Medical Hospital. A facility designed for the treatment of persons with illnesses other
than mental disorders.
Rehabilitation Unit. A residential treatment facility designed for the care of patients with
drug or alcohol problems.
Federal Prison. A confinement facility administered by the Federal government having
custodial authority over persons sentenced to confinement.
Not Known. Location where the inmate is to serve his/her sentence is not known.
Examples
An offender is sentenced to serve 5 years for a possession of marijuana conviction. Due to prison
overcrowding he is to be housed in the local jail. The correct code is "local jail."
An offender is admitted to prison to serve 5 years for a possession of marijuana conviction. She is
then placed in a drug treatment facility and will stay there through the completion of the program
- a minimum of 1 year. The correct code is "Rehabilitation Unit."
An offender is sentenced to serve 5 years for a possession of marijuana conviction. He is to serve
his sentence in a Federal penitentiary. The correct code is "Federal Prison."
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 40
Variable 18: Additional Offenses since Admission Date
Applies To
Prison Releases (Part B)
Prison Custody (Part D)
Definition
Any additional offense imposed after the date of admission (Variable 8), regardless of the date of
the crime.
Codes / Coding Information
Use your own state's offense codes
Additional Information
If, after admission, a revocation of post-confinement community supervision (PCCS) occurred
and the inmate received a sentence for violating his/her conditions of supervision, please specify
your state codes for probation or parole violation offenses as appropriate.
Examples
A parolee is readmitted to prison for violating his parole. After three months in prison he receives
an additional 5 year sentence for a new burglary conviction. The correct code is your state code
for burglary.
Variable 19: Additional Sentence Time since Admission
Applies To
Prison Releases (Part B)
Prison Custody (Part D)
Definition
The maximum time the inmate may be incarcerated consecutive to the sentence length coded in
Variable 15.
Codes / Coding Information
Report a life or a death sentence using either your agency’s codes or the following NCRP codes.
(99996)
(99997)
(99994)
(99993)
Additional sentence is Life.
Additional sentence is Death.
Additional sentence is Life plus additional years.
Additional sentence is Life without parole.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 41
Examples
An inmate assaults a guard while incarcerated and earns an additional 2 years on his remaining 7
year sentence. The correct value to report is 2 years.
A parolee is readmitted to prison for violating her parole with 6 months remaining on her
sentence. After three months in prison, she receives an additional 5 year sentence for a new
burglary conviction to be served consecutive to the current sentence. The correct value to report is
5 years.
An offender released to post-confinement community supervision is readmitted to prison for
violating conditions of supervision with 5 years remaining on her sentence. After being admitted
to prison, she receives an additional 5 year sentence for a new burglary conviction to be served
concurrent to the current sentence. The correct value to report is 0 years.
Variable 20: Prior Felony Incarcerations
Applies To
Prison Releases (Part B)
Prison Custody (Part D)
Definition
An offender who has ever been sentenced to confinement for a felony as a juvenile or adult prior
to his/her current prison admission (Variable 8).
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
(2)
(9)
Yes
No
Don’t Know
Additional Information
Do not include detention before trial or sentencing.
Do not report non-incarceration sentences such as probation, unless at some point prison time
occurred.
Examples
Ten years ago, a man served 3 years in prison for robbery and was released, having satisfied the
conditions of his sentence. He is once again being admitted to begin serving time on a new
sentence. The correct code is "Yes."
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 42
Variable 21: AWOL or Escape
Applies To
Prison Releases (Part B)
Definition
Was the offender AWOL (the failure to return from an authorized temporary absence) or did he
escape (the unlawful departure from physical custody or flight from the custody of correctional
personnel) while serving a sentence?
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
(2)
(9)
Yes
No
Unknown
Additional Information
Include in this category any inmate who escaped or was AWOL while serving time on this
sentence, regardless of whether they returned to prison or not.
Examples
An offender has completed his prison term of 5 years for larceny. During the first year of his
sentence, he escaped from prison and was returned soon thereafter. The correct value is code
"Yes."
Variable 22a: Community Release Prior to Prison Release
Applies To
Prison Releases (Part B)
Definition
Prior to release from the custody of a prison system, was the inmate concurrently under
community based supervision or placement? This includes programs such as halfway houses,
work furloughs, etc.
Examples
An inmate is admitted from prison to the state work release program on February 1st of the
reporting year. He continues to serve his sentence while working in the community. On March 1st
of the same year, he is returned to prison in order to be released. The correct value for Variable
22a is "Yes."
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 43
Variable 22b: Number of Days on Community Release
Applies To
Prison Releases (Part B)
Definition
The number of days the inmate was on community release prior to release from prison, if the
inmate was concurrently under community based supervision or placement prior to release from
the custody of a prison system.
Examples
An inmate is admitted from prison to the state work release program on February 1st of the
reporting year. He continues to serve his sentence while working in the community. On March 1st
of the same year, he is returned to prison in order to be released. The correct value for Variable
22a is "Yes.” In Variable 22b, the correct value is 28 days, the number of days on community
release prior to prison release.
Variable 23a: Date of Release from Prison
Applies To
Prison Releases (Part B)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
The most recent calendar date that the state's prison custody terminated.
Report partial dates if the day or month is not known.
Additional Information
On post confinement community supervision release (Part F) records, “Date of Release from
Prison” is the most recent prison release date prior to the post confinement community
supervision release date.
Variable 23b: Location at Time of Prison Release
Applies To
Prison Releases (Part B)
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 44
Definition
The type of facility that had been used for the custody or care of the offender just prior to release.
The name of the facility can be provided instead. In this case, provide information in a separate
file that will enable Abt Associates to re-code the name of facility into the NCRP facility type
categories listed below.
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
State Prison Facility. A confinement facility administered by the state with custodial
authority over adults sentenced to confinement.
(2)
Local Jail. A confinement facility administered by an agency of the local government,
intended for adults but sometimes also containing juveniles (holds persons detained
pending adjudication and/or persons committed after adjudication, usually with sentences
of a year or less).
(3)
Other – Specify. All facilities except those listed above which house sentenced prisoners.
Provide documentation for the types of facilities you include in this category.
(4)
Halfway House. A long-term residential facility in which residents are allowed extensive
contact with the community (e.g., attending school).
(5)
Community Work Center or Work Release. A residential facility in which residents are
employed and allowed extensive contact with the community.
(6)
Pre-release Center. A residential facility in which inmates may be placed in order to seek
employment, housing, etc.
(12)
Federal Prison. A confinement facility administered by the Federal government with
custodial authority over persons sentenced to confinement.
(99)
Unknown. Information on the facility from which the inmate is released is not known.
Examples
An offender served a 2-year prison term for burglary in the local jail due to overcrowding at the
state penitentiary. This would be coded as Local Jail.
An offender was sentenced to 18 months for a drug offense. The first 12 months were served in a
drug rehabilitation program in a county hospital. The offender then served the rest of his sentence
in prison. This would be coded as State Prison Facility.
Variable 24: Agencies Assuming Custody at Time of Prison Release
Applies To
Prison Releases (Part B)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 45
Definition
Type and location of the agency/agencies that assumes custody (physical or supervisory) over an
inmate at the time of prison release.
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(00)
None
(01)
Other Prison Outside of State
(02)
Other Prison - Federal System
(03)
Parole Within State (Include Parole Agencies in DOC)
(04)
Parole Outside State
(05)
Parole - Federal System
(06)
Probation within State
(07)
Probation Outside State
(08)
Probation Federal System
(09)
Mental/Medical Facility within State
(10)
Mental/Medical Facility Outside of State
(11)
Mental/Medical Facility - Federal
(12)
Other Within State – Specify
(13)
Other Outside State – Specify
(14)
Other - Federal – Specify
(99)
Not Known
Examples
An inmate is released from a state prison to a detainer from Federal authorities. He is transported
to a Federal prison in another state. "Other Prison, Federal" is the correct value to report.
After serving two-thirds of his sentence, an offender is required by law to be placed on mandatory
conditional release. He will be supervised by the paroling authority of that state. "Parole, Within
State" is the correct value to report.
Variable 25: Type of Release From Prison
Applies To
Prison Releases (Part B)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
Method of or reason for departure from the custody of your prison system on the reported date of
release (in Variable 23a of the current record).
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 46
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(01)
Discretionary Release Decision. A conditional release granted by a parole board or other
agency that has the authority to release adult prisoners to post-confinement community
supervision, to revoke PCCS, and to discharge an offender from PCCS.
(02)
Mandatory Conditional Release. A conditional release from prison which is mandated by
law rather than granted by a discretionary authority.
(03)
Probation Release. A conditional release to court supervision or supervision by a
probation authority after the inmate is confined usually for a brief period in a prison
facility. These cases are often called "Split Sentences" or "Shock Probation."
(04)
Other Conditional Releases – Specify. All other conditional releases not covered by the
preceding categories. Always describe the nature of the release in your documentation.
(05)
(06)
(07)
(08)
(09)
(10)
(11)
(12)
(13)
(14)
(27)
Expiration of Sentence. The termination of the period of time an offender has been
required to serve in a state prison.
Commutation/Pardon. A reduction of the term of confinement or an executive order
excusing the remainder of the sentence and pardon resulting in immediate unconditional
release.
Release to Custody, Detainer, or Warrant. Unconditionally releasing an inmate to
custody of another authority. The original prison authority relinquishes all claims upon
the inmate.
Other Unconditional Release – Specify. All unconditional releases not covered by the
preceding three categories. Always document the nature of the release.
Death by Natural Causes. Death due to illness, old age, AIDS, etc.
Death by Suicide.
Death by Homicide by Another Inmate.
Death by Other Homicide. The death of an inmate caused by a person who is not an
inmate that is not legally justifiable.
Death by Execution
Death by Other – Specify. All deaths not covered by the preceding six categories. Always
document the manner of death. Use code 14 "Other" to report an inmate's death which is
due to accidental injury caused by another person (whether the other person is an inmate
or not).
Death by Accidental Injury to Self. Death caused by the inmate accidentally injuring
himself.
(15)
Transfer. The movement of a person from the custody of your state's correctional system
to the custody of another authority while serving the same sentence. Transfers are
permanent or indefinite releases for such purposes as long-term mental health
commitment, safekeeping in another state, or housing in a Federal facility.
(16)
Release on Appeal or Bond. An offender is released to seek or participate in an appeal of
his case and is not receiving credit on his sentence while out of confinement. If the
inmate is being given credit on the remainder of his time while out of confinement or
bond, or appealing his case, do not report a release.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 47
(25)
AWOL/Escape. An inmate who is absent from your state's custody without leave or has
escaped from state prison. If your state reports AWOLs and Escapes as releases, you
must report their recapture as admissions.
(17)
Other – Specify. All other releases not specifically defined in the above categories.
Specify in your documentation the type of releases included in this category.
(99)
Not Known. The type of release from prison is not known.
Additional Information
Verify that all releases included in the Other category are releases from the custody of this prison
system and releases of sentenced persons.
For Code 16 do not include temporary movements to court (e.g., to testify or appear at a brief
hearing).
Do include transfers to other states to continue serving a sentence.
Do not include movements from prison facility to prison facility within your state.
Do not include movements of state prisoners to local jails because the prison is crowded or for
such reasons as overcrowding, safekeeping, etc.
State inmates housed in local jails are to be considered as state prison inmates.
Do not include temporary absences for such reasons as court appearances, training or medical
care.
A detainer is an official notice from one authority agency to another authority agency requesting
that a person wanted by them, but subject to the other agency's jurisdiction, not be released or
discharged without notification to the authority agency requesting the person.
The placing of a detainer is often, but not always, prior to the issuing of a warrant. Typical
reasons for the detainer are that the person is wanted for trial in the requesting jurisdiction or is
wanted to serve a sentence.
Conditional Release is the release from a federal or state correctional facility of a prisoner who
has not completed his/her sentence, and whose freedom is contingent upon obeying specified
rules of behavior while in the community. The offender can be re-incarcerated on current
sentence(s).
Persons on mandatory supervised release are usually subject to the same conditions as offenders
released to post-confinement community supervision via discretionary release, and can be
returned to prison for technical violations of release conditions. However, the difference is that
the release is not a discretionary decision of a parole board or other authority.
If you need to report a type of release not defined by one of the codes provided, assign a unique
code and define it in your documentation.
Examples
For Code 01 (Discretionary Release Decision),
An inmate is granted a release by the Parole Board after serving 3 years of a 10 year
sentence. Use code "Discretionary Release Decision."
For Code 02 (Mandatory Conditional Release),
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 48
An inmate received a 3 year sentence for heroin possession. The law requires that the
inmate be released to post-confinement community supervision after serving a year. Use
code "Mandatory Conditional Release."
For Code 03 (Probation Release),
An offender serves 180 days in prison and returns to court for a hearing. The judge
allows him to serve the remainder of his sentence on probation. The correct code is
"Probation Release."
For Code 05 (Expiration of Sentence),
A person given a maximum sentence of 5 years for robbery is released, without parole
supervision, after serving 5 years. His release is code 05, "Expiration of Sentence."
A person given a maximum sentence of 5 years for robbery is released without parole
supervision, after serving 3 1/2 years and receiving 1 1/2 years of irrevocable "Good
Time.” His release is "Expiration of Sentence."
For Code 06 (Commutation/Pardon),
After the legislature reduced marijuana offenses from felonies to misdemeanors, the 15
year sentence of a person is reduced by the Governor to actual time served, 2 1/2 years,
and the inmate is unconditionally released. The correct code is "Commutation/Pardon."
For Code 07 (Release to Custody, Detainer, or Warrant),
A man is serving three years for armed robbery in Maine. Extradition papers from Texas
on another armed robbery charge await him, however, so he is released to Texas custody.
The correct code is "Release to Custody, Detainer, or Warrant."
For Code 15 (Transfer),
An inmate is threatened by other inmates. He is transferred to the custody of another state
to complete his sentence. Use code "Transfer."
On June 10th of the report year, a Texas inmate is sent from the state prison to the
Department of Corrections training school. On June 24th of the report year, the training is
completed and the inmate is sent back to the state prison. No admission or release
movement should be reported.
Due to crowding, a Maine inmate is transferred on June 6th of the report year from the
Maine State Correctional Center to the Maine State Prison. No admission or release
movement should be reported.
An inmate is admitted to a Rhode Island prison on February 1st of the report year, to
begin serving a three year sentence for armed robbery. On June 5th of the report year, the
inmate is transferred to a county detention facility for safekeeping. No admission or
release movement should be reported.
Variable 26: Date of Release from Post Confinement Community Supervision
Applies To
Post Confinement Community Supervision Releases (Part F)
Definition
The date of discharge or termination from post-confinement community supervision for any
reason, including returning the offender to prison.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 49
Report partial dates if the day or month is not known.
Examples
An offender is discharged after completing his term of post-confinement community supervision
(PCCS) on August 1, 2008. The date of release from PCCS is August 1, 2008.
While on parole, an offender commits an armed robbery and is sentenced to serve time for that
offense. His parole is revoked, and he enters prison to begin serving time on the new sentence on
March 20, 2010. The date of release from PCCS is March 20, 2010.
Variable 27: Type of Release from Post Confinement Community Supervision
Applies To
Post Confinement Community Supervision Releases (Part F)
Definition
The reason for the termination of post-confinement community supervision that occurred on the
date provided in Variable 26.
Codes/Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(01)
(02)
(03)
(04)
(05)
(06)
Discharged, Completion of Term. The release of offenders on Post Confinement
Community Supervision (PCCS) who have served full-term sentences or who have been
released early due to a discretionary decision, commutation or pardon.
Discharged, Absconder. The release of offenders on PCCS while known to be on
absconder status, regardless of whether a warrant has been issued.
Discharged to Custody, Detainer or Warrant. Your state supervising authority or agency
relinquishes its jurisdiction over the offender on PCCS. Another agency or authority (in
or out of your state) assumes jurisdiction and perhaps custody over the person. The
agency that assumes jurisdiction or jurisdiction and custody may be a non-correctional
agency, e.g., a mental hospital.
Returned to Prison or Jail, New Sentence. The re-admission of an offender on PCCS into
a prison or jail after receiving a sentence for a new offense(s). If PCCS has been revoked
and the person is admitted to prison or jail with a new sentence, the type of release is
code 04, "Returned to Prison or Jail, New Sentence."
Returned to Prison or Jail, PCCS Revocation. The re-admission of an offender on PCCS
into a prison or jail due to the violation of the conditions of supervision, and the PCCS
has been revoked.
Returned to Prison or Jail, PCCS Revocation Pending. The re-admission of an offender
on PCCS into a prison or jail for the alleged violation of the conditions of supervision. A
revocation hearing will be held in the future and a decision to revoke or not revoke the
person's PCCS will be made.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 50
(07)
Returned to Prison or Jail, Charges Pending. The re-admission of an offender on PCCS
into a prison or jail for an alleged new offense, pending trial, conviction, or sentence.
(08)
Transferred to Another Jurisdiction. Jurisdiction over the offender on PCCS is
transferred to another state from your authority.
(09)
Death
(10)
Other – Specify. For any other removal from PCCS not covered in the previous
categories, code as 10. Please provide documentation for all PCCS exits included in this
category.
(99)
Not Known. Information on type of release from PCCS is not available.
Additional Information
Do not include those interstate compact cases where only supervisory responsibility is transferred
but legal jurisdiction is retained by your state parole authority, i.e., parole termination is still
determined by your state.
Code 02 should be used only if the offender has been formally discharged by the supervising
agency or if PCCS jurisdiction has been relinquished.
If the supervising agency changes the absconder from active to inactive status without
relinquishing jurisdiction over the person, a PCCS release should not be reported.
A detainer is an official notice from one authority agency to another authority agency requesting
that a person wanted by them, but subject to the other agency's jurisdiction, not be released or
discharged without notification to the authority agency requesting the person.
The placing of a detainer is often, but not always, prior to the issuance of a warrant. Typical
reasons for detainers are that the offender is wanted for trial in the requesting jurisdiction.
If an offender on PCCS has had his supervision status revoked because he violated conditions
of supervision but was not sentenced for another crime, code as 05, "Returned to Prison or Jail,
PCCS Revocation."
Count persons returned to prison or jail with revocation pending only if termination of PCCS
jurisdiction is pending in your state.
Use one of the codes 04-07 for absconders who have been released from PCCS because he was
returned to jail or prison.
For parolees who have already received new sentences at the time of release from PCCS, code as
04, "Returned to Prison or Jail, New Sentence."
Count persons returned to prison or jail with charges pending.
Examples
For Code 01 (Discharged, Completion of Term),
A parolee, released from prison, is required to serve three years on parole. He finishes the
three years and is discharged by the Adult Parole Authority. Use code 01, "Discharged,
Completion of Term."
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 51
An offender, released from prison, is required to serve three years on mandatory
conditional release. He finishes two years and receives an early discharge by the
supervising agency. Use code 01, "Discharged, Completion of Term."
For Code 02 (Discharged, Absconder),
Wyoming parolee moved to New Mexico last year without the permission of the
Wyoming Board of Parole. After six months, the Wyoming Parole Board relinquished
jurisdiction. Use code 02, "Discharged, Absconder."
An offender on post-confinement community supervision in Nevada moved to New
Mexico last year without permission of the Nevada supervising agency. As of December
31 of the report year, the Nevada supervising agency had not relinquished jurisdiction.
Do not submit a PCCS exit record for this offender.
For Code 03 (Discharged to Custody, Detainer or Warrant),
A Wisconsin probationer is discharged as a result of an extradition request from Texas.
He is released to Texas custody on a warrant. Use code 03, "Discharged to Custody,
Detainer or Warrant."
For Code 04 (Returned to Prison or Jail, New Sentence),
While out on supervised release, an offender commits a crime and is sentenced to serve
two years in prison. PCCS is revoked. Use code 04, "Returned to Prison or Jail, New
Sentence."
For Code 05 (Returned to Prison or Jail, PCCS Revocation),
A probationer in Wisconsin violates the conditions of his probation. The supervising
agency formally revokes his probation and the offender is returned to the county jail to
continue serving his sentence. Use code 05, "Returned to Prison or Jail, PCCS
Revocation."
For Code 06 (Returned to Prison or Jail, Revocation Pending),
A parolee is accused of violating conditions of his parole. He is sent to the state prison to
await a decision from the Parole Authority concerning possible revocation. Use code 06,
"Returned to Prison or Jail, Revocation Pending."
For Code 07 (Returned to Prison or Jail, Charges Pending),
An offender on supervised release is charged with committing a new offense. He is held
in the local jail to await trial on the new charge. Use code 07, "Returned to prison or jail,
charges pending."
For Code 08 (Transferred to Another Jurisdiction),
A parolee in Mississippi finds a new job in Alabama. The Mississippi Parole Board
arranges for the parolee to be supervised in Alabama through an interstate compact
agreement. Your state parole authority has not relinquished jurisdiction; therefore no
parole exit has occurred.
An offender on PCCS in Mississippi finds a new job in Alabama. The Alabama Board of
Pardons and Paroles agrees to assume jurisdiction over the parolee; Mississippi then
terminates jurisdiction. Use code 08, "Transferred to Another Jurisdiction."
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 52
Variable 28: Supervision Status Just Prior to Release
Applies To
Post Confinement Community Supervision Releases (Part F)
Definition
Level of contact during the year prior to release from post confinement community supervision.
Codes/Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(01)
Active. Include persons required to make contact (in person, by mail, or telephone) with
the supervising authority at least once a month during the last year of post confinement
community supervision (PCCS).
(02)
Inactive. All offenders on PCCS who were excused from reporting on a regular basis
during the last year of PCCS supervision but were held accountable and remained under
your agency's jurisdiction.
(03)
Absconded. Any offender on PCCS who has not been discharged but fails to report to the
supervising authority, as was instructed, or who leaves the geographical area of
supervision without permission.
(04)
Supervised Out of State. Any offender whose PCCS is supervised by a state other than
yours but your state retains jurisdiction of the offender.
(05)
Other – Specify. For any offender on PCCS who had a supervision status just prior to
release not covered by the above categories, code as 05. Please document the nature of
their supervision status.
(06)
Only have financial obligations remaining.
Additional Information
Include both active and inactive cases as defined above.
Examples
A parolee visits his parole officer the first Friday of every month. Use code 01, "Active."
A probationer receives a form once a month in the mail from his probation officer. He completes
it and sends it back. Use code 01, "Active."
An offender has been on supervised release for five years. After three years of active supervision,
no active contact is required. Use code 02, "Inactive."
A Wyoming parolee moves to New Mexico without the permission of the Wyoming Parole
Board. Parole jurisdiction is soon relinquished. Use code 03, "Absconded.” If Wyoming does not
relinquish jurisdiction, no parole exit should be reported to NCRP.
An Arizona offender on post-confinement community supervision finds a new job in Texas. The
Texas Board of Pardons and Parole agrees to monitor his supervision although the Arizona
supervising agency does not relinquish jurisdiction. Supervision is terminated by Texas when
Arizona terminates the offender supervision. This PCCS release should be reported by Arizona as
code 04, "Supervised Out of State."
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 53
(There is no Variable 29)
Variable 30: Inmate State ID Number
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
The inmate's unique, fingerprint-supported State Identification (SID) Number assigned by the
state’s criminal history repository.
Additional Information
All information that can identify individuals will be held strictly confidential by Abt Associates
and the Bureau of Justice Statistics as required by Title 42, United States Code, Sections 3735
and 3789g.
Variable 31a: Indeterminate Sentence
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Definition
Is any part of the total maximum sentence reported in Variable 15 an indeterminate sentence (a
sentence in which the judge specifies a minimum and maximum prison term)?
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
(2)
(9)
Yes
No
Don’t Know
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 54
Examples
An offender is serving a 10-year determinate sentence for robbery under a truth in sentencing law,
and a 5-year sentence for drug trafficking under a mandatory minimum law.
“No” for variable 31a (indeterminate sentence),
“Yes” for variable 31b (determinate sentence),
“Yes” for Variable 31c (mandatory minimum sentence), and
“Yes” for variable 31d (restricted by a truth in sentencing law).
An offender is serving a 10 to 15-year indeterminate sentence for vehicular homicide, a 5-year
determinate sentence for reckless endangerment, and a 3-year determinate sentence for driving
under the influence of drugs. The 10 to 15-year indeterminate sentence for vehicular homicide is
restricted by a truth in sentencing law. The vehicular homicide sentence is not a mandatory
minimum, nor is the 5-year sentence for reckless endangerment. It is not known whether the 3year sentence for driving under the influence of drugs is a mandatory minimum sentence. The
correct entry is:
Variable 31a (indeterminate sentence) –Yes.
Variable 31b (determinate sentence) –Yes.
Variable 31c (mandatory minimum) – Not Known.
Variable 31d (truth in sentencing) – Yes.
Variable 31b: Determinate Sentence
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Definition
Is any part of the total maximum sentence reported in variable 15 a determinate sentence (a
sentence in which the judge sets a fixed prison term)? The sentence may be reduced by good time
credits or earned time.
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
(2)
(9)
Yes
No
Don’t Know
Examples (see Variable 31a)
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 55
Variable 31c: Mandatory Minimum Sentence
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Definition
Is any part of the total maximum sentence reported in variable 15 a mandatory minimum sentence
(a minimum sentence specified by statute for a particular crime)?
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
(2)
(9)
Yes
No
Don’t Know
Examples (see Variable 31a)
Variable 31d: Truth in Sentencing Restriction
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Definition
Is any part of the total maximum sentence reported in Variable 15 restricted by a Truth in
Sentencing Law (a statute which mandates that a certain percentage of the court-imposed
sentence be served in prison)?
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
Yes
(2)
No
(9)
Don’t Know
Examples (see Variable 31a)
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 56
Variable 32: Length of Court-Imposed Sentence to Community Supervision
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Definition
The amount of time which the court states that the offender is required to serve under community
supervision after release from prison.
Additional Information
This variable is applicable only if the court imposed a sentence to community supervision that is
separate from the sentence to prison.
The sentence to post-incarceration community supervision may be in the form of parole,
probation, or other supervision in the community, as ordered by the court.
Examples
The offender is sentenced by the court to serve a 5-year fixed prison term and an additional 2year term on community supervision after release from prison. The correct value to report is 2
years.
The offender is sentenced by the court to serve a 2 to 10-year sentence in prison. The court did
not sentence the offender to a separate term of community supervision. The term of community
supervision will be determined by an administrative agency, such as a parole board, when the
offender is approved for release from prison. The correct value to report is “not applicable.”
Variable 33: Parole Hearing / Eligibility Date
Applies To
Prison Admissions (Part A)
Prison Custody (Part D)
Definition
The date the offender is eligible for review by an administrative agency such as a parole board, to
determine whether he or she will be released from prison.
Report partial dates if the day or month is not known.
Additional Information
This variable is applicable only if the decision to release an offender is controlled by an
administrative agency such as a parole board.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 57
The parole hearing eligibility date should be calculated from the total maximum sentence
(variable 15) for all offenses. For the year-end custody record, report the next date the inmate will
be eligible for a parole hearing.
Examples
An offender was admitted to prison on January 1, 1999, with a 15 years to life sentence for
second degree murder. The law states the offender is eligible for parole board release after
serving 85% of the minimum 15-year sentence (or 12 years 9 months). The parole eligibility date
is calculated by adding 12 years 9 months to the date of admission. The offender will be eligible
for parole board release on October 1, 2011.
A judge sentences an offender to serve 2 to 4 years in prison for theft. The offender is eligible for
parole board release after the minimum 2-year sentence has been served. The offender was
admitted to prison on January 1, 2010, with 6 months in jail time credits. The parole eligibility
date is calculated by adding two years to the date of admission, and subtracting six months for
credited jail time. The parole eligibility date is July 1, 2010.
An offender is admitted to prison on January 1, 2005, with a 10-year sentence for aggravated
robbery. The law requires violent offenders to serve 50% of the sentence before they are eligible
for parole board release. Good time credits may be accrued only after 50% of the sentence has
been served. The parole eligibility date is 5 years from the date of admission, or January 1, 2010.
While on parole, an offender is arrested for aggravated assault and is sentenced to a 10-year
prison term for the new offense. At sentencing, the offender’s parole is revoked with 2 years
remaining on a previous robbery sentence. The offender is admitted to prison on January 1, 2004
as a parole violator, with a 12-year total maximum sentence for both convictions. Good time
credits may be accrued only after 50% of the sentence has been served. The parole eligibility date
is 6 years from the date of admission, or January 1, 2010.
Variable 34: Projected Release Date
Applies To
Prison Admissions (Part A)
Prison Custody (Part D)
Definition
The projected date on which the offender will be released from prison.
Report partial dates if the day or month is not known.
Additional Information
Statutory requirements, good time credits, jail time credit, and any other factors which might
modify the prison release date should be included in this calculation.
If an offender is serving time for more than one offense, the projected release date should be
calculated from the total maximum sentence (variable 15) for all offenses.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 58
Examples
An offender enters prison on January 1, 2002, with a 10-year sentence for armed robbery. At
sentencing, the offender received 6 months credit for time served in jail prior to being admitted to
prison. While in prison, the State allows inmate to earn one day work credit for every 3 days
served, not to exceed 15% of the sentence. The projected release date is calculated by subtracting
the 6 months jail credit and the 1 ½ years of available work credit from the 10-year prison
sentence. The offender’s projected release date is 8 years from the date of admission or January 1,
2010.
A judge sentences an offender to serve 10 years in prison for armed robbery. The offender is
admitted to prison on January 1, 2002, and is required by State law to serve 6/7 of the 10-year
sentence (8.57 years, or 8 years 6 months and 26 days). The offender’s projected release date is 8
years 6 months and 26 days from the date of admission or July 26, 2010.
A judge sentences an offender to serve 2 to 6 years in prison for theft. The offender is admitted to
prison on January 1, 2007, and is given 3 years of good time credit (one-half the maximum
sentence). Assuming the offender does not lose any good time while incarcerated, he or she is
projected to be released after serving the remaining 3 years of the maximum sentence. The
projected release date is calculated as January 1, 2010.
A judge sentences an offender to serve 5 to 10 years in prison for aggravated robbery. The
offender is admitted to prison on January 1, 2000, and given 5 years of good time credit (one-half
the maximum sentence). After serving 8 years the offender has lost all good time credits due to
disciplinary actions. The offender is expected to expire the sentence, or serve the entire 10-year
maximum sentence, and release unconditionally from prison. The projected release date is 10
years from the date of admission or January 1, 2010.
While on parole, an offender is arrested and convicted for armed robbery and sentenced to a 10year prison term for the new offense. The offender’s parole is revoked with 2 years remaining on
a pervious robbery sentence. The offender is admitted to prison on January 1, 2004 as a parole
violator, with a 12-year total maximum sentence for both robbery convictions. The offender is
given 6 years of good time credit at admission (one-half the total maximum sentence). The
projected release date is 6 years from the date of admission, or January 1, 2010.
Variable 35: Mandatory Release Date
Applies To
Prison Admissions (Part A)
Prison Custody (Part D)
Definition
The date the offender by law must be conditionally released from prison.
Report partial dates if the day or month is not known.
Additional Information
This date should reflect jail time credits and any statutory or administrative sentence reductions,
including good time.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 59
The mandatory release date should be calculated from the total maximum sentence (variable 15)
for all offenses.
This variable is intended to capture mandatory conditional release policies structured around good
time and other administrative sentence reductions.
Do not set to the date the offender’s sentence will expire (serve the entire sentence and be
released unconditionally from prison).
Examples
An offender is admitted to prison on January 1, 2006, with a 5 to 10-year prison sentence for
fraud. The law requires mandatory release for non-violent offenders when good time credits plus
actual time served in prison equals the maximum sentence. The offender is allowed to earn a
maximum of 45 days good time credit for every 30 days served. The mandatory release date is
calculated by determining the date the offender’s actual time served plus good time will equal the
maximum sentence. After serving 4 years, the offender will have earned a maximum of 6 years in
good time credit. The mandatory release date is 4 years from the date of admission, or January 1,
2010.
Variable 36: First Name
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
The first name of the offender.
Additional Information
All information that can identify individuals will be held strictly confidential by Abt Associates
and the Bureau of Justice Statistics, in accordance with Title 42, United States Code, Sections
3735 and 3789g.
Variable 37: Last Name
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 60
Definition
The last name of the offender.
Additional Information
All information that can identify individuals will be held strictly confidential by Abt Associates
and the Bureau of Justice Statistics, in accordance with Title 42, United States Code, Sections
3735 and 3789g.
Variable 38: Facility Name
Applies To
Prison Custody (Part D)
Definition
Name of the facility in which the prisoner will be incarcerated at yearend.
Variable 39: FBI Number
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
The unique identification number given by the Federal Bureau of Investigation/ Interstate
Identification Index to each offender.
Codes / Coding Information
All information that can identify individuals will be held strictly confidential by Abt Associates
and the Bureau of Justice Statistics as required by Title 42, United States Code, Sections 3735
and 3789g.
Variable 40: Prior Military Service
Applies To
Prison Admissions (Part A)
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 61
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
Did the inmate ever serve in the U.S. Armed Forces?
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
(2)
(9)
Yes. Does not require that the inmate receive veterans’ benefits, nor that the inmate
served in a conflict situation. Includes all branches of the military, including the Coast
Guard.
No
Don’t Know
Variable 41: Date of Last Military Discharge
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
The date the inmate was discharge from the U.S. Armed Forces for the final time.
Report partial dates if the day or month is not known.
Variable 42: Type of Last Military Discharge
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 62
Definition
The type of discharge the offender received from the U.S. Armed Forces on the date in Variable
41.
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(9)
Honorable. Offender received a rating from good to excellent for their service.
General (honorable conditions). Offender’s military performance was satisfactory.
General (not honorable conditions). Offender’s military performance was satisfactory but
marked by a considerable departure in duty performance and conduct expected of
military members.
Other than honorable. Offender’s military performance was a serious departure from the
conduct and performance expected of all military members.
Bad conduct. Only given by a court martial.
Dishonorable. May be rendered only by conviction at a general court-martial for serious
offenses that call for dishonorable discharge as part of the sentence.
Other.
Not Known.
Variable 43: Date of Admission to Post Confinement Community Supervision
Applies To
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
The date the offender was most recently admitted to post-confinement community supervision on
the current sentence.
Report partial dates if the day or month is not known.
Variable 44: Type of Admission to Post Confinement Community Supervision
Applies To
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
The reason an offender entered into post-confinement community supervision on the date
provided in Variable 43 (Date of Admission to Post-Confinement Community Supervision) of the
current record.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 63
As necessary, provide information in a separate file that will enable Abt Associates to re-code
your agency’s PCCS admission type codes into the NCRP PCCS admission type categories listed
below.
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1)
(2)
(3)
(4)
(5)
(6)
(9)
Discretionary release from prison. An offender being admitted to PCCS based on the
decision of the Governor, the department of correction, or parole board, or commutation
of sentence.
Mandatory conditional release from prison. An offender being admitted to PCCS based
on a determinate sentencing statute or good-time provision
Reinstatement of PCCS. Offenders returned to PCCS status, including discharged
absconders whose cases were reopened, revocations with immediate reinstatement, and
offenders re-admitted to PCCS at any time under the same sentence.
Court-imposed sentence to PCCS that begins upon release from prison. An offender
being admitted to PCCS based on a judicial sentence of a period of incarceration
immediately followed by a period of PCCS.
Transferred from another jurisdiction. An offender admitted following a term of
confinement or community supervision in another state when that state transfers legal
authority of the offender to your state.
Other.
Not known.
Variable 45: County Where Offender was Released / County Where PCCS
Office is Located
Applies To
Post Confinement Community Supervision Releases (Part F)
Definition
The county where the offender was released from post-confinement community supervision on
the date in Variable 26.
If this information is not available, please report the county where the post-confinement
community supervision (PCCS) office to which the offender reported before exit is located.
Codes / Coding Information
If possible, use either the name of the county or the 5-digit county FIPS code (available at
http://www.itl.nist.gov/fipspubs/co-codes/states.txt).
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 64
Variable 46: Social Security Number
Applies To
Prison Admissions (Part A)
Prison Releases (Part B)
Prison Custody (Part D)
Post Confinement Community Supervision Admissions (Part E)
Post Confinement Community Supervision Releases (Part F)
Definition
The 9-digit number assigned by the U.S. Social Security Administration to indicate a unique
individual.
If this information is not available or your state does not allow the reporting of full 9-digit SSN,
please report the last 4 digits of SSN.
Codes / Coding Information
Please do not include dashes.
Variable 47: Street Address of Residence Prior to Imprisonment
Applies To
Prison Admissions (Part A)
Post Confinement Community Supervision Admissions (Part E)
Definition
Text field allowing for as much of the recorded street address as available for an offender’s last
known residence prior to imprisonment
Codes / Coding Information
Please include all street numbers, apartment numbers, housing units, etc. if possible.
Variable 48: City of Residence Prior to Imprisonment
Applies To
Prison Admissions (Part A)
Post Confinement Community Supervision Admissions (Part E)
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 65
Definition
City of an offender’s last known residence prior to imprisonment
Variable 49: State of Residence Prior to Imprisonment
Applies To
Prison Admissions (Part A)
Post Confinement Community Supervision Admissions (Part E)
Definition
State of an offender’s last known residence prior to imprisonment
Variable 50: Zip Code of Residence Prior to Imprisonment
Applies To
Prison Admissions (Part A)
Post Confinement Community Supervision Admissions (Part E)
Definition
5-digit zip code of an offender’s last known residence prior to imprisonment
Codes / Coding Information
Please do not include dashes
Variable 51: Custodial Security Level
Applies To
Prison Custody (Part D)
Definition
Security level at which an offender is held during imprisonment
Codes / Coding Information
Use either your agency’s codes or the following NCRP codes for this variable.
(1) Maximum/close/high custody - assigned to prisoners requiring the highest degree of
supervision because they pose a danger to others and to the institution; or because their well-
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 66
being would be in jeopardy if they refused protective custody. These prisoners cannot
participate in activities requiring outside movement, and their inside movement is closely
observed.
(2) Medium custody - assigned to prisoners needing more than minimal supervision. Their inside
movement and call-outs require passes and/or supervision. Their outside movement requires
restraints except for work or program assignments.
(3) Minimum/low custody - assigned to prisoners posing the least threat to the institution and
public safety. They include inmates assigned to community service centers and halfway
houses and those who participate in work, education, and other activities in the community.
They are generally permitted to move unescorted for program and work assignments.
(4) Not classified/other – Inmates are unsentenced or sentenced and awaiting classification
(5) Unknown.
Abt Associates Inc.
2015 NCRP Data Request Instructions ▌pg. 67
Appendix I
NCRP frequently asked questions fact sheet
National Corrections Reporting Program (NCRP) FAQs
January 2015
What is the National Corrections Reporting Program (NCRP)?
NCRP compiles offender-level data on admissions to and releases from prisons and post-confinement community
supervision programs. The Bureau of Justice Statistics (BJS) has administered the NCRP since 1983. State
departments of correction and community supervision provide these data, which are used at the federal and state
levels to monitor correctional populations and address policy questions related to recidivism, prisoner reentry, and
trends in demographic characteristics of the incarcerated and paroled populations.
What is the Bureau of Justice Statistics?
The Bureau of Justice Statistics (BJS), a component of the Office of Justice Programs in the U.S. Department of
Justice, is the United States' primary source for criminal justice statistics. Its mission is to collect, analyze, publish,
and disseminate information on crime, criminal offenders, victims of crime, and the operation of justice systems at
all levels of government. These data are critical to federal, state, and local policymakers in combating crime and
ensuring that justice is both efficient and evenhanded.
How many states participate in NCRP?
Last year 48 states submitted NCRP data. Our goal is 100% participation.
What is Abt Associates’ role in NCRP?
Abt Associates was awarded a grant in October 2010 by the Bureau of Justice Statistics to direct the NCRP. (Prior
to that date, the U.S. Census Bureau was the NCRP data collection agent.) Abt is responsible for collecting,
processing and analyzing data submitted by state departments of corrections and community supervision. Working
with BJS, Abt will also implement BJS’s vision of an enhanced and expanded NCRP system that provides timely
and useful information to federal and state policymakers.
What is Abt Associates?
Abt Associates is a global leader in research and program implementation in the fields of social and economic
policy, health, and international development. Abt Associates has 40 years of experience working for the U.S.
Department of Justice and criminal justice agencies across the country. Known for its rigorous approach to solving
complex challenges, Abt Associates is regularly ranked as one of the top 20 global research firms. The employeeowned company has multiple offices in the U.S. and program offices in nearly 40 countries.
What data is collected under NCRP?
State departments of correction are asked to submit three data files:
Prison Admissions (Part A): one record for each admission of a sentenced offender to the state’s
prison system.
Prison Releases (Part B): one record for each release of a sentenced offender from the state’s
prison system.
Prison Custody (Part D): one record for each sentenced offender in the physical custody of the
state’s prison system at year end.
State agencies responsible for supervising offenders on a term of community supervision immediately after
release from prison are asked to submit two data files:
Post Confinement Community Supervision Admissions (Part E): one record for each admission to a
post-confinement community supervision program.
Post Confinement Community Supervision Releases (Part F): one record for each release from a
post-confinement community supervision program.
Most states submit these data these data annually, with the submissions containing admissions and releases
from the previous calendar year.
What data elements are requested in these files?
The data elements differ somewhat across the five data files, but generally include:
Offender characteristics (e.g., unique agency identifier, name, date of birth, race, sex, veteran
status)
Sentence characteristics (e.g., county where sentence imposed, offenses, sentence length)
Date and type of admission to prison
Date and type of release from prison
Date and type of admission to post-confinement community supervision
Date and type of release from post-confinement community supervision
The NCRP data request documentation contains complete information on all the requested data elements.
What if we are unable to provide all of these data elements?
If your agency does not collect one or more of the requested data elements or providing them would be an
excessive burden (or is not allowed under agency policy), those data elements do not have to be included in the
data submission. The data request documentation also highlights the “core” data elements that are most important
to NCRP.
How long will it take us to respond to this data request?
The amount of time depends on the characteristics of your agency’s offender information system, the type of data
extraction tools available for that system, and the level of expertise agency staff have in using those tools. The
largest time commitment is in the first year of participation, when data extract procedures must be developed. BJS
estimates the time needed to develop computer programs to extract data and to prepare a response to be 24 hours,
on average, per type of database containing the information needed, for the first year of participation, and 8 hours,
per type of database, during the second and subsequent years. Feedback during data processing and review is
estimated to take 2 hours. Send comments regarding this burden estimate or any other aspects of the collection of
this information, including suggestions for reducing this burden, to the Director, Bureau of Justice Statistics, 810
Seventh Street, NW, Washington, DC 20531, and to the Office of Management and Budget, OMB number 11210065, Washington, DC 20503. For more information on the NCRP reporting burden (OMB No. 1121-0065 Exp.
1
10/31/2015), see the NCRP's OMB submission.
When is the data submission due?
The target date for submitting NCRP data is March 31 of each year, but we understand that agency constraints in
many states preclude meeting that target date. The Abt NCRP site liaison assigned to your state will work with you
to set a realistic target date.
Is there a specific format or coding scheme for the data?
There is no required format or coding scheme for the data you submit.
1
http://www.reginfo.gov/public/do/PRAViewICR?ref_nbr=201208-1121-005
How do we submit the NCRP data?
The preferred method for submitting data to Abt Associates is via the NCRP data transfer site
(transfer.abtassoc.com). This site is compliant with FIPS (Federal Information Processing Standard) 140-2 and
meets all the requirements of the Federal Information Security Management Act (FISMA) and the Privacy Act. The
data are automatically encrypted during transit.
How can we be assured that data we submit is secure?
BJS and Abt are bound by federal law (42 USC 3789g) which provides that, “No officer or employee of the Federal
Government, and no recipient of assistance under the provisions of this chapter shall use or reveal any research or
statistical information furnished under this chapter by any person and identifiable to any specific private person for
any purpose other than the purpose for which it was obtained in accordance with this chapter. Such information and
copies thereof shall be immune from legal process, and shall not, without the consent of the person furnishing such
information, be admitted as evidence or used for any purpose in any action, suit, or other judicial, legislative, or
administrative proceedings.” Abt further recognizes that it is bound by the Privacy Act and the Federal Information
Security Management Act (FISMA) regarding how NCRP data are received, processed, and released.
What happens after we submit data?
Abt will verify the contents of the data files and conduct a series of validity checks on the data, including comparing
the submitted data to your submissions from prior years. Typically, this will be accomplished within 2-4 weeks of
receipt of your data. Your Abt site liaison will then contact you to review our findings. Having a thorough
understanding of what data you submit is necessary in order to construct valid and reliable national NCRP datasets.
How will the data be used?
NCRP data are intended to be used at the federal and state levels to address policy questions related to recidivism,
prisoner reentry, and trends in demographic characteristics of correctional and community supervision populations.
BJS uses NCRP data to monitor these issues at the national level. Abt Associates actively solicits ideas from state
NCRP contacts on how NCRP data can be used in their state. Researchers at universities and other institutions
can access NCRP data – minus offender unique identifiers and names – at the National Archive of Criminal Justice
Data (http://www.icpsr.umich.edu/icpsrweb/NACJD/), following a review by an Institutional Review Board (IRB).
Who do we contact for more information?
Tom Rich (Abt Associates Project Director and site liaison) – tom_rich@abtassoc.com or 617-349-2753
Michael Shively (Abt Associates site liaison) – michael_shively@abtassoc.com or 617-520-3562
Ann Carson (BJS Program Manager) – elizabeth.carson@ojp.usdoj.gov or 202-616-3496
Or, visit www.ncrp.info
Appendix J
Examples of follow‐up emails to 5 states seeking clarification on NCRP data submitted
Carson, Elizabeth
From:
Sent:
To:
Subject:
Tom Rich
Thursday, April 16, 2015 3:54 PM
Carson, Elizabeth
questions for North Carolina
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
‐‐‐‐‐Original Message‐‐‐‐‐
From: Tom Rich
Sent: Tuesday, April 07, 2015 11:29 AM
To: 'Stevens, Cara'
Subject: RE: 2014 NCRP data request
Hi Cara,
I hope you enjoyed the conference. Good seeing you again.
Thanks again for submitting the 2014 NCRP data. We've processed the data and have a few questions for you, to make
sure we're understanding the data correctly.
1. Education (variable 7) is a value from 0 to 20. Does that correspond to the number of grades completed?
2. The jurisdiction on date of admission (variable 10) is missing in 99% of cases. In previous years this variable has been
almost entirely "North Carolina". This year should we treat missing as "North Carolina"?
3. We see lots of offenses listed in the records. Should we treat the first one in the record as the most serious /
controlling offense?
4. Many of the offenses are listed more than once in a particular record. Should we assume that the offense count is
equal to the number of occurrences of a particular offense code within a record?
5. In the type of admission to parole variable (variable 44), what does "North Carolina case" mean?
6. In the type of release from parole variable (variable 27), what does "unsupervised" mean?
7. In the type of release from parole variable (variable 27), does "unsatisfactory termination" mean they were revoked
back to prison?
8. Are the variables MSCMTMAX, PLMAXSNT, and CMTRMPRB in the form YYYMMDD?
9. The variable "prior felony" (variable 20) is "No" in all but 2 of the Part D (custody) records. In 2013, it was 54% yes
and 46% no. For this year, should we treat this variable as missing?
Let me know if you need more clarification on these questions.
Thanks,
Tom
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
1
‐‐‐‐‐Original Message‐‐‐‐‐
From: Stevens, Cara [mailto:cara.stevens@ncdps.gov]
Sent: Monday, March 30, 2015 9:55 AM
To: Tom Rich
Subject: RE: 2014 NCRP data request
Thanks Tom.
Cara Stevens, M.A.
Correctional Research & Evaluation Analyst Rehabilitative Programs & Services Division of Adult Correction and Juvenile
Justice North Carolina Department of Public Safety
Phone: (919) 324‐6488
Fax: (919) 715‐7754
cara.stevens@ncdps.gov
www.ncdps.gov
‐‐‐‐‐Original Message‐‐‐‐‐
From: Tom Rich [mailto:Tom_Rich@abtassoc.com]
Sent: Monday, March 30, 2015 9:26 AM
To: Stevens, Cara
Cc: Edwards, David
Subject: Re: 2014 NCRP data request
Thanks Cara! We received all the files. We'll process the files in the next few days and let you know if we have any
questions.
Tom
Tom Rich
Abt Associates Inc.
617‐349‐2753
> On Mar 30, 2015, at 9:22 AM, Stevens, Cara wrote:
>
> Tom,
>
> I uploaded the North Carolina datasets this morning.
>
> Thanks,
> ‐Cara
>
> Cara Stevens, M.A.
> Correctional Research & Evaluation Analyst Rehabilitative Programs &
> Services Division of Adult Correction and Juvenile Justice North
> Carolina Department of Public Safety
> Phone: (919) 324‐6488
> Fax: (919) 715‐7754
> cara.stevens@ncdps.gov
> www.ncdps.gov
>
>
2
> From: Tom Rich [mailto:Tom_Rich@abtassoc.com]
> Sent: Wednesday, March 25, 2015 9:09 AM
> To: Edwards, David
> Subject: RE: 2014 NCRP data request
>
> David,
>
> Great news that the data are ready! Thanks very much.
>
> I'm in the process of getting you login credentials for the NCRP file transfer site. I'll call you (hopefully this morning; if
not, Friday afternoon) once I have them. In the meantime, I've attached the general instructions for using the site.
>
> Talk to you soon.
>
> Thanks,
> Tom
>
> Tom Rich | Senior Associate | Abt Associates
> 55 Wheeler St. | Cambridge, MA 02138
> O: 617.349.2753 | F: 617.492.5219 |
> www.abtassociates.com
>
> From: Edwards, David [mailto:David.Edwards@ncdps.gov]
> Sent: Tuesday, March 24, 2015 6:09 PM
> To: Tom Rich
> Subject: RE: 2014 NCRP data request
>
> Hi Tom: I believe we are ready to upload the data. If you would like to give me a call at your convenience tomorrow
before 1:00 with the log‐in information, I'll be available. Otherwise, I'll be in the office on Friday after 1:00 as well.
Thanks, David.
>
> David Edwards, MRP
> Policy Development Analyst
> Rehabilitative Programs & Services
> Division of Adult Correction & Juvenile Justice NC Department of
> Public Safety Mail Service Center 4221
> 3040 Hammond Business Place
> Raleigh, NC 27699‐4221
> Phone 919.324.6480
> Fax 919.715.7754
> david.edwards@ncdps.gov
> www.ncdps.gov
>
> From: Tom Rich [mailto:Tom_Rich@abtassoc.com]
> Sent: Tuesday, March 24, 2015 10:18 AM
> To: Edwards, David
> Subject: 2014 NCRP data request
>
> David,
>
3
> I just wanted to check back with you on the NCRP data request. Do you think you'll be able to submit the data within
the next couple weeks? Our informal deadline is March 31st, and your agency has always been able to meet that date in
the past.
>
> Thanks,
> Tom
>
> Tom Rich | Senior Associate | Abt Associates
> 55 Wheeler St. | Cambridge, MA 02138
> O: 617.349.2753 | F: 617.492.5219 |
> www.abtassociates.com
>
> From: Tom Rich
> Sent: Monday, January 12, 2015 9:24 AM
> To: Edwards, David
> (David.Edwards@ncdps.gov)
> Subject: 2014 NCRP data request
>
> Hello David,
>
> I've attached the official request for 2014 NCRP data, as well as the instructions and an FAQ.
>
> I know this is your first year as point of contact for NCRP, so please contact me if you have any questions. I'm
assuming Pam left solid documentation on how to run the extract programs that she developed. You'll be glad to hear
that there aren't any changes to the data request from last year.
>
> On behalf of BJS, thank you very much for your support of NCRP.
>
> Tom
>
> Tom Rich | Senior Associate | Abt Associates
> 55 Wheeler St. | Cambridge, MA 02138
> O: 617.349.2753 | F: 617.492.5219 |
> www.abtassociates.com
>
>
> ________________________________
> This message may contain privileged and confidential information intended solely for the addressee. Please do not
read, disseminate or copy it unless you are the intended recipient. If this message has been received in error, we kindly
ask that you notify the sender immediately by return email and delete all copies of the message from your system.
>
> ________________________________
>
> E‐mail correspondence to and from this address may be subject to the North Carolina Public Records Law and may be
disclosed to third parties by an authorized state official.
>
> ________________________________
> This message may contain privileged and confidential information intended solely for the addressee. Please do not
read, disseminate or copy it unless you are the intended recipient. If this message has been received in error, we kindly
ask that you notify the sender immediately by return email and delete all copies of the message from your system.
>
4
Carson, Elizabeth
From:
Sent:
To:
Subject:
Tom Rich
Thursday, April 16, 2015 3:57 PM
Carson, Elizabeth
Kentucky questions (first round, anyway).
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
From: Tom Rich
Sent: Tuesday, February 10, 2015 2:34 PM
To: 'Hall, Johnathan (DOC)'
Cc: 'Moore, Beth (DOC)'
Subject: RE: 2014 NCRP data request
John,
It has taken us a long time to get to your 2014 NCRP, but we finally got to it today. The Part A records are the
same ones submitted with the 2013 data last year. Looks like your ‘year’ parameter was updated for all the
Parts except Part A.
Tom
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
From: Hall, Johnathan (DOC) [mailto:Johnathan.Hall@ky.gov]
Sent: Monday, January 12, 2015 4:14 PM
To: Tom Rich
Subject: RE: 2014 NCRP data request
You are quick ‐ I didn’t even have time to finish an email telling you they had been uploaded!
Please let us know if you need anything further.
Thanks,
John
From: Tom Rich [mailto:Tom_Rich@abtassoc.com]
Sent: Monday, January 12, 2015 4:13 PM
To: Hall, Johnathan (DOC)
Subject: RE: 2014 NCRP data request
John,
1
I see all 5 files there. Many thanks!
We are in the middle of preparing our annual data submission to BJS, so we probably won’t be able to review
these files for a week or two. But I will get back to you and Beth if we have any questions on the files.
Thanks again,
Tom
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
From: Hall, Johnathan (DOC) [mailto:Johnathan.Hall@ky.gov]
Sent: Monday, January 12, 2015 4:08 PM
To: Tom Rich
Subject: RE: 2014 NCRP data request
As sure and I am typing this email, I just tried the exact same credentials as I did earlier (several times) and they
worked. I am in now.
Sorry for the trouble!
From: Tom Rich [mailto:Tom_Rich@abtassoc.com]
Sent: Monday, January 12, 2015 4:06 PM
To: Hall, Johnathan (DOC)
Subject: RE: 2014 NCRP data request
I was just able to log in with those credentials. Double check that you entered colemanc for the user
name. All lower case, although I don’t think case matters on the user name. Ampersand is &. Also, just to
make sure, the pound sign is #.
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
From: Hall, Johnathan (DOC) [mailto:Johnathan.Hall@ky.gov]
Sent: Monday, January 12, 2015 4:02 PM
To: Tom Rich
Subject: RE: 2014 NCRP data request
Tom –
I attempted to log on with the user name ‘colemanc’ and the password you provided. The site is giving me the following
message “Invalid username/password or not allowed to sign on from this location.”
Just to make sure I am correct – when you referred to the ampersand you meant the “&” symbol, correct? After there
any letters in the username that should be capitalized?
2
From: Tom Rich [mailto:Tom_Rich@abtassoc.com]
Sent: Monday, January 12, 2015 3:52 PM
To: Hall, Johnathan (DOC)
Subject: RE: 2014 NCRP data request
John,
Let me know if you have any trouble uploading. I get a notification when the files land, so I’ll let you know
when they’re all here.
Thanks.
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
From: Hall, Johnathan (DOC) [mailto:Johnathan.Hall@ky.gov]
Sent: Monday, January 12, 2015 3:47 PM
To: Tom Rich
Cc: Moore, Beth (DOC)
Subject: RE: 2014 NCRP data request
Tom –
I don’t mind at all to use Cedric’s account. Please call me at (502) 782‐2257.
Thanks,
John
From: Tom Rich [mailto:Tom_Rich@abtassoc.com]
Sent: Monday, January 12, 2015 2:31 PM
To: Hall, Johnathan (DOC)
Cc: Moore, Beth (DOC)
Subject: RE: 2014 NCRP data request
John,
Wow – you guys are fast!
The upload instructions are attached. We have an account under Cedric Coleman’s name – user name
colemanc. The password is one that I created so I’m ok with your using that account, if you’re ok with
it. Otherwise I’ll create an account for you or Beth.
Let me know what you’d prefer. Either way, I have to tell you the password over the phone, so also let me
know a number to reach you at.
Thanks,
Tom
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
3
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
From: Hall, Johnathan (DOC) [mailto:Johnathan.Hall@ky.gov]
Sent: Monday, January 12, 2015 2:23 PM
To: Tom Rich
Cc: Moore, Beth (DOC)
Subject: RE: 2014 NCRP data request
Hi Tom –
I hope you are doing well.
As you probably gathered from her out of office message, Beth is on vacation this week. In her absence, I have prepared
the data extract for you and all five files are ready for submission. However, I do not have any credentials or information
about how to access your FTP site. If you wouldn’t care to provide that information I will be happy to upload the files for
you.
Thanks!
John
From: Tom Rich [mailto:Tom_Rich@abtassoc.com]
Sent: Monday, January 12, 2015 10:39 AM
To: Hall, Johnathan (DOC)
Cc: Moore, Beth (DOC)
Subject: 2014 NCRP data request
Hi Johnathan,
I hope you are doing well, and that the new year is off to a good start.
It’s that time of the year when we contact states to request NCRP data for the prior year (2014). I have attached the
data request letter, a letter of support from BJS, the instructions, and an FAQ. You and Beth will be glad to hear that
there aren’t any changes to the data request from last year.
I hope we will see you at the NCRP meeting in Colorado in a couple months.
Thanks,
Tom
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
This message may contain privileged and confidential information intended solely for the addressee. Please do
not read, disseminate or copy it unless you are the intended recipient. If this message has been received in error,
we kindly ask that you notify the sender immediately by return email and delete all copies of the message from
your system.
4
Carson, Elizabeth
From:
Sent:
To:
Subject:
Tom Rich
Thursday, April 16, 2015 3:55 PM
Carson, Elizabeth
questions for Nevada DOC
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
From: Tom Rich
Sent: Tuesday, March 17, 2015 9:43 AM
To: 'Alejandra Livingston'
Subject: RE: 2014 File Upload
Hi Alejandra,
Thank you again for submitting the 2014 NCRP file. We’ve reviewed the file, and had a couple questions for
you.
1. What do the offense codes 00A007, 00A008, 00A009, and 00A010 mean? The offense description for
these codes is “Aggregate”.
2. Based on our discussions in January, we were expecting to see the new ID field we requested last year (the
NDOC number) at the end of each record. But we didn’t see a new field at the end of each record. Was the
NDOC number included in the files? We see an ID field in each record, but we assume that was the inmate ID
that you’ve been providing for the past several years.
Thanks,
Tom
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
1
From: Alejandra Livingston [mailto:alivingston@doc.nv.gov]
Sent: Friday, March 06, 2015 10:55 AM
To: Tom Rich
Cc: Andrea Franko; Dwayne Deal
Subject: 2014 File Upload
Good morning Tom, this message is to advise you that Nevada has submitted the 2014 NCRP files via the file transfer
wizard provided by your firm.
Feel free to contact me should you have any questions.
Regards,
Alejandra C. Livingston, MS
Research, Planning, & Statistics
Nevada Department of Corrections
P.O. Box 7011
Carson City, NV 89702
Ph: (775) 887‐3357
Fax:(775) 887‐3243
(Please note that my fax number has changed)
This preceding e‐mail message and accompanying documents are covered by the Electronic communications Privacy Act, 18
U.S.C. SS 2510‐2521, and contain information intended for specific individuals(s) only or constitute non‐public
information. This information may be confidential. If you are not the intended recipient you are hereby notified that you have
received this document in error and that any review, dissemination, copying, or the taking of any action based on the contents
of this information is strictly prohibited. If you have received this communication in error, please notify me immediately by e‐
mail, and delete the original message. Use, dissemination, distribution or reproduction of this message by unintended
recipients is not authorized and may be unlawful.
This message may contain privileged and confidential information intended solely for the addressee. Please do
not read, disseminate or copy it unless you are the intended recipient. If this message has been received in
error, we kindly ask that you notify the sender immediately by return email and delete all copies of the
message from your system.
2
Carson, Elizabeth
From:
Sent:
To:
Subject:
Tom Rich
Thursday, April 16, 2015 3:54 PM
Carson, Elizabeth
questions for Pennsylvania parole
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
From: Tom Rich
Sent: Thursday, March 05, 2015 11:30 AM
To: wimoser@pa.gov
Cc: Klunk, Frederick (fklunk@pa.gov)
Subject: NCRP 2000-2014 EFG files
Bill,
Many thanks again for uploading the 2000‐2014 EFG files. We really appreciate your willingness to go back
that far. We have a few questions for you below, to make sure we understand the data.
1. Does the variable CountyResidence refer to the county where the offender is living while on parole? where
he was living when the sentence was imposed? where the court that sentenced the offender is located?
2. There are a few records that have a date of admission to parole in the year 3209. Should we set these
values to “missing”?
3. There are a few race codes that appear infrequently in the data that we don’t know the meaning of (as
necessary, we can set these to “missing”):
0
C
P
S
M
4. What does EntryCode 4A mean?
5. There are a few codes for Sex that we don’t know the meaning of:
0
7
B
E
U
6. There are a few StatusCode that we don’t know the meaning of:
40
1
41
42
43
44
45
7. There are about 65 offense codes that we don’t know the meaning of (as necessary, we can set these to ‘unknown –
these codes are rarely used)
05
25
28
140
213
215
216
219
226
235
236
239
246
253
255
262
270
272
277
282
283
284
285
288
290
293
441
853
863
942
20M
51I
55I
6OO
96O
A46
A91
A99
Aid
All
2
B15
B19
B22
COR
Cre
Dea
Dis
DRI
Fle
GRA
I.D
Inv
Man
Par
PIC
Pos
Pro
PWI
R.E
REC
Sim
STA
Sto
Str
TER
Una
Thanks again,
Tom
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
This message may contain privileged and confidential information intended solely for the addressee. Please do
not read, disseminate or copy it unless you are the intended recipient. If this message has been received in error,
we kindly ask that you notify the sender immediately by return email and delete all copies of the message from
your system.
3
Carson, Elizabeth
From:
Sent:
To:
Subject:
Tom Rich
Thursday, April 16, 2015 3:55 PM
Carson, Elizabeth
questions for Pennsylvania DOC
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
From: Tom Rich
Sent: Tuesday, April 14, 2015 2:13 PM
To: Flaherty, Robert (rflaherty@pa.gov)
Subject: 2014 NCRP files
Hi Bob,
(Now on to our other survey….). Thanks again for submitting the 2014 NCRP files. We’ve reviewed the files
and have a couple questions for you:
1. We noticed that the total maximum sentence length is a date. I just wanted to confirm that to calculate the
maximum sentence length in years and month, we should subtract the date of admission to prison from the
maximum sentence length date in the files.
2. There are a few offense codes that we don’t know the meaning of. Here they are (if it’s easier to send us a
more recent offense code table, please do so):
7533091
185112A
184303A
187613A
184955A2
187615A4
236311A
342522A
424137A1
7533001A1
425947F
CC00010
CC182714
CC25007D
CC2603A
CC2701A4
1
CC2709A6
CC2906
CC3125A5
CC3307A1
CC4303
CC4955
CC551121
CC5511H3
CC5516
CC6141
CPV
HIGH
LOW
JC4134
JC9712.1
PC43002
TCV
TPV
VC3736A
VOIP
Thanks,
Tom
Tom Rich | Senior Associate | Abt Associates
55 Wheeler St. | Cambridge, MA 02138
O: 617.349.2753 | F: 617.492.5219 | www.abtassociates.com
This message may contain privileged and confidential information intended solely for the addressee. Please do
not read, disseminate or copy it unless you are the intended recipient. If this message has been received in error,
we kindly ask that you notify the sender immediately by return email and delete all copies of the message from
your system.
2
Appendix K
Whitepaper on estimating the proportion of the prison population that will serve long sentences using
NCRP data
“Estimating Prison Stays Among Current Prison Populations”
Jeremy Luallen, Ph.D., Abt Associates, Inc.
Gerry Gaes, Ph.D., Florida State University
Chris Cutler, M.A., Abt Associates, Inc.
DRAFT: PLEASE DO NOT CITE WITHOUT PERMISSION FROM THE AUTHORS
February 23, 2015
This work was supported by Grant No. 2010-BJ-CX-K067 awarded by the Bureau of Justice
Statistics, Office of Justice Programs, U.S. Department of Justice. For this work, Thomas Rich
served as Project Director along with Principal Investigators William Rhodes and Gerry Gaes.
Points of view in this document are those of the authors and do not represent the official position
of the U.S. Department of Justice. The authors are responsible for any errors in the paper.
Abstract
A recent report by the Pew Center describes the impact of increasingly longer prison terms on
the costs of corrections over the past two decades. In that report, the author’s forecast the
expected length of stay for the current prison population based on prison exit rates (i.e., prison
stock divided by exits in a single year) and use a common “release cohort” approach to estimate
the average length of stay. While these approaches can provide reasonable estimates of prison
stays lengths, they have drawbacks in application. This paper proposes an alternative method for
estimating the length of stay among prison populations using a survival model with left
truncation. Its express intent is to forecast the number of short-stay, medium-stay and long-stay
prisoners in the current population. We argue that this approach offers several advantages which
make it a better tool for corrections officials to forecast the number of short-stay, medium-stay
and long-stay prisoners. Prison authorities can use this forecast to allocate resources more
efficiently.
Introduction
Measuring length of stay in prison is an important metric for criminal justice researchers.1 It is
used to explain changes in prison populations over time and describe variations in punishment
across jurisdictions. It is also used to measure equity and proportionality in sentencing across
individuals. Traditionally, estimates of time-served focus on the fundamental question: “What is
the average time an offender serves in prison?” However, there are practical benefits to knowing
the proportion of the population that will remain in prison for short terms as compared to long
terms. This suggests an alternative question: “Of all the offenders currently in prison, how many
are expected to serve sentences of specified lengths?” This alternative question has been raised
by researchers (Pew, 2012) and government agencies (the Bureau of Justice Statistics) and
expresses a different empirical objective – measuring the number offenders by length of stay.
Though the answer to this question also appears to be important, it has been historically
underemphasized in the literature, likely because strong assumptions are required to produce
reasonable estimates.
At present, methods appearing in the literature for estimating prison stays in current prison
populations are limited. Estimates based on aggregate prison stocks and flows describe the
average expected stay, but do not describe the distribution. Estimates based on release cohorts
describe the distribution, but are highly variable, require additional data about sentencing and do
not provide confidence intervals. Moreover, they require strong assumptions about the flow of
prisoners both into and out of prison over time. In light of these limitations, we discuss a new
method for estimating the expected length of stay of current prison populations. Specifically we
propose an estimator that uses a survival model with left truncation and right-hand censoring to
estimate the distribution of projected length of stay and then use these estimates to quantify the
size of short, medium, and long term offender groups. While this survival method also requires
some strong assumptions, it has advantages in application that make it a superior choice for
estimation.
We test the robustness of our proposed estimator using data from the National Corrections
Reporting Program (NCRP). Overall we find that estimates derived through our alternative
approach are an improvement over estimates obtained from release cohorts. Estimates of shortstay offenders, where variability matters least, show comparability between methods. Estimates
of long-stay offenders, where variability matters most, show notably less volatile but otherwise
reasonable estimates using a survival-based approach. In addition, estimates using our approach
require fewer assumptions and no special treatment of offenders with life sentences. Ultimately
we argue that a survival-based approach to estimation offers a better tool for forecasting the
number of short-stay, medium-stay and long-stay prisoners. The remainder of this paper is
organized as follows. First we motivate our method by describing its importance to practitioners.
1
This is, of course, distinct from the sentence an offender receives. Actual time served can be quite different from
the associated sentence, especially in jurisdictions without sentencing guideline procedures
Second we describe the competing methods: the release cohort method and the survival method.
Next we describe the data we use for our estimations, followed by a comparison of results using
each method. Finally, we offer some concluding remarks.
Motivation
From a practical perspective, knowing the sizes of current offender populations by projected
length of stay is a benefit to corrections administrators and practitioners. The reason is that
decisions made by corrections officials about the day-to-day administration of corrections are
directly impacted by prisoner composition according to stay length. Consider, for example,
differences in prisoner socialization and prison culture that occur with the balance of long-stay
and short-stay inmates. Some ethnographies of prisoner life discuss the cultural tone set by lifers
who “just want to do their time” versus inmates who have relatively short duration terms
(Clemmer, 1940; Irwin, 2009; Sykes, 2007). While there has not been a great deal of research on
the inmate composition with respect to length of stay, Toch and Adams (1989) found that
inmates with long lengths of stay were less likely to commit misconduct than those with short
lengths of stay. They also found that misconduct was more likely to occur at the beginning of a
term and the probability declined over time. Other research has described aspects of
“prisonization” for long-term inmates that include changes in socialization and thought patterns
(Wilson & Vito, 1988). To the extent that the mix of prisoner types is a predictor of prisoner
behavior, knowing this mix helps decision makers to optimize correctional staff allocation,
implement routines that maximize prison stability and introduce policies that promote
correctional objectives.
Alternatively, knowing the number of long-stay (vs. short-stay) inmates promotes the effective
allocation of prison budgets. For example, current and planned allocation of prison health care
dollars may depend on the size and distribution of long-stay inmates. Many long-stay inmates
are naturally older, placing greater fiscal strain on corrections budgets (Wilson & Vito, 1988;
Chettiar, Bunting & Schotter 2012; Fellner & Vinck 2012). Similarly, to the extent that the
programming and treatment needs of long-stay prisoners are different from short and mediumstay prisoners, optimal allocation of those dollars may vary with the mix of prisoner types. As
the proportion of long-stay prisoners grows, administrators may wish to devote more resources to
the promotion of coping strategies and related activities for which long-stay inmates are
receptive, or to offering more intensive and targeted pre-release preparation (Adams, 1992;
Wilson & Vito, 1988).
Setting aside their practical benefits, the motivation for these estimates is also partly driven by
the same normative considerations that lead policymakers and researchers to measure time
served in the first place. To the extent that measuring average length of stay is seen as useful,
measuring the number of offenders should be equally useful. Both measures serve to improve
the public’s understanding and help public officials to make informed policy decisions – goals
unto themselves. Given the objective, our method leads to better estimates of length of stay and
improves the utility of this measure.
Data
For this exercise we use data from the National Corrections Reporting Program (NCRP),
operated by the Bureau of Justice Statistics. The design of this dataset has important
implications for our proposed model, so we describe it in some detail here. Moreover, the NCRP
data are the same data used by Pew in their earlier report (2012).
The NCRP is a longitudinal file that tracks individual offenders within prison populations over
time. The time frame covered for this longitudinal file in many states is 2000 to 2013. Over this
time frame, the NCRP records information about (a) every offender admitted to prison regardless
of when they were released, (b) every offender released from prison regardless of when they
were admitted, and (c) every offender appearing in prison at some point regardless of when they
were admitted. The implication of this design is that we can observe outflows for offenders
admitted decades in the past and covering selected windows of time. It is this abundance of
longitudinal data that makes the NCRP a rich source of data for analysis of this type. Moreover
it is a public-use dataset, providing a platform for others to conduct similar analysis.
In addition to offender-level data on prison admissions, releases and stocks, the NCRP also
collects other important data elements including sex, date of birth, race, offense, and sentence
length. While we do not exploit these additional data elements for this paper, they offer potential
multivariate extensions to the survival modeling we propose here. We restrict our dataset to 38
states in all. In these states, data have been transformed into a longitudinal format, are known to
have been tested and certified for reliability and extend to December 31, 2013.
Methods
As described earlier, the Pew report (2012) estimates expected length of stay using a stock-flow
ratio. Expected time-served is computed as the ratio of the prison stock to the flow of releases
during the year. Such an approach is known to the field (Blumstein and Beck 1999; Blumstein
and Beck, 2005; Patterson and Preston, 2008) and has appeal in that it is a simple computation
with minimal data requirements. A disadvantage of this estimator is that it leads to an estimate
of the mean but does not estimate the distribution of prisoners by length of stay. Where the
objective is to forecast the size of groups by length of stay, more extensive use of release cohorts
is required. In the following text, we describe an approach for estimating counts using release
cohorts, followed by a description of our proposed survival method.
Release Cohort Method
The logic behind using release cohorts to forecast the distribution of stay length is
straightforward and requires two assumptions. The first is that prisoners with length of stay S are
admitted and released at a constant rate. That is to say that offenders admitted in year 1 are
released in year (1 + S), offenders admitted in year 2 are released in year (2 + S), and so on. This
method must also assume that admission groups are of equal size. With these assumptions, the
size of a group (with stay S) can be directly estimated based upon the observed releases in any
given year, since the number of releases (i.e., admissions) will be constant over time. The
estimate is multiplicative to the number of releases in a given year with stay length S. For
example, the estimated number of offenders with stay, S = 3 in a given population is just 3 times
the number of released offenders of S = 3 in a given year.
Equations [1] and [2] below generalize this condition. Let 𝑁𝑇𝑆 be the estimated stock population
of offenders with stay length S in year T. Also let 𝑅𝑇𝑆 be the number of offenders with observed
length of stay S released in year T, such that the estimated stock is:
[1]𝑁𝑇𝑆 = 𝑆 ∗ 𝑅𝑇𝑆
By simple extension, the estimated overall stock population (in year T) is just the summation of
𝑁𝑇𝑆 over every S:
[2]𝑁𝑇 = ∑𝑆𝑖=0 𝑆 ∗ 𝑅𝑇𝑆
There some notable features of this model that limit its utility. First, because estimated stocks
are built entirely from release cohorts, they ignore observable information about the current
prison population. To the extent that admission rates in fact vary over time, estimated stocks
need not resemble observed stocks. As an illustration, consider an extreme example where a
state admits enough offenders during the last year so that their population doubles. Because the
majority of new offenders will not have been released by the end of the year, their presence in
any release cohort is not yet observed. The influx of new offenders is unaccounted for and
estimates will understate the number of offenders in the stock by close to half. This
complication drives the need for the steady-state assumption about admissions.
Second, significant variability is driven by sentences that are both very long and very rare. The
reason is that small groups of offenders are represented at a rate that is multiplicative to their stay
length. Consider for example a release cohort that includes one offender released after 35 years.
By its construction, this formula implies that there are 1*35 = 35 similar offenders in the stock.
However, if by chance this release cohort contained two such offenders, now the estimate
number of similar offenders doubles to 2*35 = 70. The result is that lumpiness in release cohort
groupings drives noisiness in stock estimates.
Third, the model does not by itself provide a formal confidence interval for estimated stocks. It
appears the only way to assess variability would be through repeated estimation across multiple
release cohorts, provided data are sufficient. Finally, the model does a poor job of handling life
sentences in many cases. There are two reasons. First, the use of life sentences has not been
uniform over time, accelerating over the past few decades (citations). Second, these increased
admissions are largely not (yet) observed in most release cohorts. Together, these factors
invalidate the steady-state assumption and, without a correction, make estimates for the longest
stays unreliable. A reasonable solution is to identify offenders with life sentences from the
current stock, then supplement estimates with these offenders as a group with a predefined length
of stay, e.g., > 20 years. This trumps the need to estimate the length of stay for these offenders
and still allows the analyst to classify an offender as being in a long term length of stay group.
This is the solution we adopt for this paper. To do so requires additional data, some of which is
not available for some states. More importantly, a method using survival modeling can handle
all of these issues.
Proposed Survival Method
In this paper, we propose to use a survival function to estimate the distribution of the length of
stay for a given prison population and describe expected group membership. A survival function
is especially useful because estimates at each interval are conditional on time. It can therefore
easily be applied to a selected prison stock where time already served is known. It also has the
advantages that confidence intervals are obtained directly from the estimation, and that life
sentences are naturally handled by the model. Moreover, estimates can be made less sensitive to
chance variations in the longest stays by setting an upper bound on the estimation. For example,
offenders with sufficiently long stays (e.g. stays of 20 years or more) can be collapsed into a
single, larger group for estimation. The release cohort method does not allow for this collapsing
because it relies on the exact stay length to draw inferences.
There are two main issues to consider for estimating and applying the survival function in this
context. The first concerns the existence of time trends in survival estimates (i.e., trends in timeserved). As a practical matter, the existence of time trends in survival estimates make it
impossible to construct reasonable estimates of the future; too much uncertainty exists. This is
equally true with both release cohort and survival methods. As such, our model must assume
survival estimates are time invariant. The assumption is strong, but the data provide a test.
Unlike the release cohort method, there is no additional need to assume constant admission rates.
The estimated survival function only uses information for offenders at risk, at the time they are at
risk. No assumptions need be made about the number of admissions, and so the survival
function itself can be applied to any number of prisoners. Because estimates are applied to the
stock population itself, projected estimates will exactly equal the number of known offenders.
The second consideration for this model is that the data are both left truncated and right-hand
censored. Censoring exists because some offenders remain in prison past the end of the
observation window (Dec. 31, 2013). Truncation occurs because offenders admitted prior to the
start of the data collection window (i.e., 2000) are only observed if they are still in prison
beginning in 2000; offenders both admitted and released before 2000 are never observed.
Truncation implies that admission cohorts before 2000 only provide partial information about the
regions of the underlying survival function they represent. For example, a 1996 admission
cohort does not provide information about survival rates in year 1, year 2, year 3 or year 4 by its
construction, but does provide information about survival rates in year 5, conditional on survival
up to that point. In fact, estimates for this cohort are identified for up to a 12-year span, from
year 5 to year 17. Estimates beyond year 17 are not supported because of censoring. Most
importantly, the presence of both censoring and truncation does not preclude reliable estimates.
Parametric and nonparametric survival models can be reliably estimated in these circumstances,
provided risk sets are appropriately constructed and sufficiently sized and other assumptions are
met (Kaplan & Meier, 1958; Woodroofe, 1985; Tsai, 1987; Chao & Lo, 1988; Gijbels & Wang,
1993).
Survival estimates for this paper are derived from a nonparametric maximum likelihood
estimator (Kaplan and Meier, 1958), shown in equation [3] below. We use a nonparametric
estimator for simplicity, though parametric estimators can also be used.
[3] Sˆ (t )
n kj d kj
nk
j |t j t
j
Equation [3] expresses the standard nonparametric (KM) estimator proposed by Kaplan and
Meier (1958), with one notational difference. As with the standard KM estimator, the subscript j
associated with risk sets n and losses d indicate events recorded as of time t j . In addition, the
superscript k indicates the data come from a subset of admission cohorts (A). The steady-state
assumption is important to the construction of this estimator and implies no adverse effects result
from the use of late entrants. Offenders across admission cohorts can be thought of as random
selections from an underlying distribution of time-served.
The construction of this k subset depends on j and can be defined in a variety of ways. The
simplest construction of k is to use only the most recent admission cohort with available data,
i.e., k ϵ {A2014-j}. In that case, estimates for year 1 survival ( the interval 0 ≤ t < 1) are derived
from the 2013 admission cohort, estimates for year 2 survival (1 ≤ t < 2) are derived from the
2012 admission cohort, and so on. This is the construction we use for estimation in this paper,
although our definitions are not limiting. This construction allows us to capitalize on the most
recent data available for a given survival period. Others may wish to expand the cohorts used in
estimation, where survival estimates are identified; however we argue that restricting estimates
to use the most recent data (i.e., the last identified cohort) provide the strongest guard against
possible trends in time-served.
Estimates of the survival curve derived from [3] above are used to estimate current stocks
according to equation [4], shown below:
T
T
[4] N | T T * G i Sˆ j
i 0
j i
Sˆ Sˆ (t j )
where Sˆ j is defined such that: j
ˆ
S j 1
for j T *
*
for j T
In this equation N | T T * is the estimated number of offenders in prison for at least T* years.
The term Gi denotes the observe stock of offenders already incarcerated for i years. The term Sˆ j
denotes the relevant portion of the survival curve applied to offender stocks Gi. Simply put,
survival estimates are applied on the range from i to T*, not including T*. All offenders
projected to be remaining in prison as of T* are simply pooled together and added to the total.
This paper treats T* as 20 years, though others may wish to apply different cutoffs.
Results
We compare estimates from our survival method to those derived from the release cohort method
using data from 38 states. Results are reported in Table 1. We stratify estimates according to
three offender groupings: short-stay, medium-stay and long-stay offenders. Short stays are
defined as prison stays of less than 5 years. Medium stays are defined as prison stays between 5
and 19 years. Long stays are defined as prison stays of 20 years or more. This stratification is
arbitrary; analysts can choose strata suitable to their requirements.
This table shows several important results. First, it shows that estimates for short-stay offenders
are extremely close between methods. In 33 of the 38 states, predicted differences are 10% or
less. In 18 states the absolute difference in predicted group sizes is less than 500 offenders, and
in 29 states this difference is less than 1,000. Estimates for medium-stay offenders are similarly
comparable, with a clear exception of estimates in California. In California, estimates based on
release cohorts dramatically overstate the prison population by over 24,000 offenders (roughly
150% of the estimated population according to the survival method). The difference is
attributable to the Realignment reforms recently adopted in California, which have dramatically
reduced admissions and subsequent stock in California prisons, but which are not accounted for
by the release cohort estimates. The result highlights the weakness of the release cohort in
instances where admission rates are significantly changing over time.
Table 1 also shows estimates for long-stay offenders between methods are generally close,
though the disparity of estimates varies substantially by state. In general, it appears that smaller
states have notably larger estimates of long-stay offenders using the survival method relative to
the release cohort method. Given the greater sensitivity of release cohort estimates to changes
among smaller populations, we argue that survival estimates provide a better prediction. Figure
1 illustrates this point. It shows the distribution of states (on the y-axis) according to the
predicted proportion of the population that is long-stay (on the x-axis). The figure compares
methods by showing the predicted distribution from the survival methods on the left, and the
same distribution for the release cohort method on the right. It shows that results from the
survival method are normally distributed, while the results for the release cohort method appear
significantly more variable. Given that these same distributions among short and medium-stays
tend toward the normal distribution for both the survival and release cohort methods, the result
on the left is arguably more credible.
Finally, we offer three figures which summarize the information from Table 1 as simple-to-read
graphics. All figures summarize estimates between methods. Figure 2 shows estimates for
short-stay offenders, Figure 3 for medium-stay, and Figure 4 for long-stay offenders. Readers
should note that the ranges of the y-axes vary between graphics.
Conclusion
The future is of course uncertain. Nevertheless, forecasting the future is important because it
enables public administrators to make decisions based upon current knowledge. In corrections,
administrators that can effectively project the sizes and attributes of their offender populations
can foresee budgetary pressures more easily and make more informed decisions about current
resource allocation. With that goal in mind, this paper proposes a method that is more reliable
than release cohort estimates of the number of inmates classified by length of stay in prison.
Using data from 38 states, we show that offender groups can be reliably estimated with a
survival model that has several advantages in application over estimation performed through
release cohorts. Overall, we argue this approach offers a better tool for forecasting.
The method we describe here is a general approach and applicable to a wide variety of
correctional settings. However, there may be specific circumstances where forecasts are better
achieved through other means. Consider, for example, states that have adopted determinate
sentencing laws. There should be much less ambiguity in these states about how long most
offenders will remain in prison. The exercise may be as simple as counting up the number of
prisoners with already known length of stays. More likely, the exercise would involve some
mixture of counting and estimation. In some states, even with determinate sentencing guideline
procedures, there is variability in length of stay because of judges’ ability to depart from
guidelines and because some states have generous good time credits. In any case, our method
provides new flexibility in how projections can be achieved.
We recognize certain limitations to the proposed method. For one, we make no distinction
between offenders serving a revocation term and a new court commitment. For the purposes of
allocating resources, this make no difference even though there is intense interest in this
distinction among criminologists and policy administrators (National Research Council, 2014).
We have also found a great deal of ambiguity in making this distinction between new court
commitments and revocations in many states that contribute to NCRP. Second, while we have
focused on the length of stay estimates without covariates, certain jurisdictions will want to use a
parametric estimation method to provide insight into the drivers associated with length of stay.
We used the Kaplan-Meier (KM) survival estimator because it requires the fewest assumptions
and given the density of the NCRP data, it provides a reasonable estimate of the survival
function. In some instances, simply stratifying the sample and applying the KM estimator will
provide sufficient insight of the effect of some limited set of covariates.
Finally, while acknowledging that the KM estimator is less sensitive to changes in length of stay
over time than the release cohort method, we acknowledge there will be jurisdictions where
unanticipated dramatic changes in length of stay will occur. Many times a policy change will be
incremental as it was in the federal system where implementation of sentencing guidelines
occurred gradually because it only applied to offenders whose crimes occurred after the date of
implementation (Gaes, Simon, and Rhodes, 1992). There will be unusual cases such as the one
that occurred in the California prison system where the shock was the result of a federal court
intervention forcing the state to send prisoners to local jails who previously had been
incarcerated in the state prison system. No method can anticipate such shocks.
Prison length of stay has implications for many elements of criminal justice investigations, both
practical and theoretical. We argue that using a left truncated right censored survival estimator
gives us the best result.
References
Adams, K. (1992). Adjusting to prison life. Crime and justice, 275-359.
Blumstein, A., & Beck, A. J. (2005). Reentry as a transient state between liberty and
recommitment (Vol. 56). New York, New York: Cambridge University Press.
Blumstein, A., & Beck, A. J. (1999). Population growth in US prisons, 1980-1996. Crime. &
Just., 26, 17.
Chao, M. T., & Lo, S. H. (1988). Some representations of the nonparametric maximum
likelihood estimators with truncated data. The Annals of Statistics, 661-668.
Clemmer, D. (1940). The Prison Community. New York, NY: Holt, Rinehart and Winston.
Chettiar, Inimai M. and Bunting, William and Schotter, Geoffrey, (2012) At America's Expense:
The Mass Incarceration of the Elderly, American Civil Liberties Union, NYU School of
Law, Public Law Research Paper No. 12-38; NYU Law and Economics Research Paper
No. 12-19.
Fellner, J., & Vinck, P. (2012). Old Behind Bars: The Aging Prison Population in the United
States. Human Rights Watch
Gaes, G. G., Simon, S, E, & Rhodes, W. M. (1992) FEDSIM: A Sentencing impact and prison
population projection model for the federal criminal justice system, Unpublished
manuscript, Federal Bureau of Prisons.
Gijbels, I., & Wang, J. L. (1993). Strong representations of the survival function estimator for
truncated and censored data with applications. Journal of Multivariate Analysis, 47(2),
210-229.
Irwin, J. (2009). Lifers: Seeking Redemption in Prison. New York, NY: Routledge.
Kaplan, E. L., & Meier, P. (1958). Nonparametric estimation from incomplete observations.
Journal of the American statistical association, 53(282), 457-481.
National Research Council (2014) The growth of incarceration in the United States : Exploring
causes and consequences / Committee on Causes and Consequences of High Rates of
Incarceration, Jeremy Travis and Bruce Western, editors, Committee on Law and Justice,
Division of Behavioral and Social Sciences and Education. National Research Council of
the National Academies.
Pew Center on the States, & United States of America. (2012). Time Served: The High Cost,
Low Return of Longer Prison Terms.
Patterson, E. J., & Preston, S. H. (2008). Estimating mean length of stay in prison: Methods and
applications. Journal of Quantitative Criminology, 24(1), 33-49.
Sykes, G. M. (2007). The society of captives: A study of a maximum security prison. Princeton
University Press.
Toch, H., Adams, K., & Grant, J. D. (1989). Coping: Maladaptation in prisons. New
Brunswick, NJ: Transaction.
Tsai, W. Y., Jewell, N. P., & Wang, M. C. (1987). A note on the product-limit estimator under
right censoring and left truncation. Biometrika, 74(4), 883-886.
Wilson, D. G., & Vito, G. F. (1988). Long-term inmates: Special needs and management
considerations. Fed. Probation, 52, 21.
Woodroofe, M. (1985). Estimating a distribution function with truncated data. The Annals of
Statistics, 163-177.
Tables
Table 1: Estimated Current State Prison Populations, by Stay and Method
Short-Stays
Medium-Stays
Survival
Release
Cohort
Diff.
AL
CA
13,163
55,071
13,281
52,707
CO
11,125
DE
Long-Stays
Survival
Release
Cohort
Diff.
Survival
Release
Cohort
Diff.
-118
2,364
8,260
43,148
6,020
67,269
2,241
-24,121
6,506
34,701
6,413
35,651
93
-950
11,910
-786
6,061
9,189
-3,128
2,235
4,159
-1,924
3,302
3,229
73
1,330
1,429
-99
750
70
680
FL
49,026
48,189
837
33,554
33,196
357
18,243
17,920
324
GA
28,935
28,731
204
16,697
16,251
447
8,037
10,522
-2,485
IN
18,862
17,428
1,434
7,762
6,457
1,304
2,853
1,700
1,153
IA
5,701
5,416
285
1,594
2,262
-668
911
75
836
KS
5,133
4,886
246
2,724
2,481
243
1,699
199
1,500
KY
15,216
16,803
-1,586
4,504
5,212
-708
1,747
1,211
535
ME
1,382
1,281
102
468
335
133
294
107
186
MA
4,521
5,140
-618
3,040
3,593
-553
1,866
629
1,238
MI
17,551
16,490
1,062
16,410
18,169
-1,759
9,291
9,647
-355
MN
6,856
6,877
-21
1,708
2,785
-1,078
622
590
33
MS
11,993
12,002
-8
6,866
5,539
1,327
3,173
2,348
825
MO
18,645
17,749
897
8,094
9,225
-1,131
4,693
3,953
740
MT
1,570
1,680
-111
577
680
-104
283
24
260
NE
3,229
2,730
499
1,275
1,158
117
496
598
-102
NV
6,951
6,338
612
3,867
3,151
716
1,923
3,133
-1,209
NH
1,663
1,600
63
690
817
-127
288
23
266
NJ
14,056
15,428
-1,373
5,660
8,490
-2,830
2,354
4,120
-1,766
NM
4,241
4,075
166
1,858
1,427
431
733
54
679
NY
28,900
26,773
2,126
15,380
16,563
-1,183
8,364
11,917
-3,553
NC
17,910
17,473
436
13,155
16,201
-3,045
5,592
4,299
1,293
ND
1,220
1,220
1
249
178
71
100
61
39
OH
28,183
27,990
193
14,190
14,627
-436
10,280
7,817
2,464
OK
13,087
11,930
1,158
8,957
8,161
797
4,516
3,762
754
OR
7,392
6,891
501
6,143
7,064
-920
1,435
3,833
-2,398
PA
29,646
26,266
3,380
13,353
15,019
-1,666
8,460
7,766
694
RI
1,828
1,656
172
528
415
113
288
235
53
SC
9,640
9,128
512
7,912
8,982
-1,071
4,314
2,641
1,673
TN
16,668
15,836
832
8,840
7,721
1,119
7,209
3,213
3,997
TX
81,490
72,400
9,090
50,198
55,546
-5,347
21,706
21,869
-162
UT
4,171
3,653
517
2,669
798
1,871
2
-
2
WA
9,894
9,195
699
5,345
5,210
134
2,426
2,698
-272
WV
4,430
4,712
-282
1,701
1,712
-11
785
303
483
WI
12,333
11,937
396
5,649
6,755
-1,106
4,175
1,573
2,602
WY
1,498
1,415
83
668
347
321
143
82
61
Total
566,484
542,446
24,037
331,084
370,433
-39,349
183,497
175,211
8,285
Figures
Figure 1: Distribution of States According to the Estimated Proportion of Long-Stay Prisoners
Release Cohort Method
30.0%
27.5%
25.0%
22.5%
20.0%
17.5%
15.0%
12.5%
10.0%
7.5%
5.0%
2.5%
0.0%
30.0%
27.5%
25.0%
22.5%
20.0%
17.5%
15.0%
12.5%
10.0%
7.5%
5.0%
2.5%
10
9
8
7
6
5
4
3
2
1
0
0.0%
Number of States
Survival Method
Figure 2: Estimated Number of Current Short-Stay Prisoners
90,000
70,000
60,000
50,000
40,000
30,000
20,000
10,000
-
AL
CA
CO
DE
FL
GA
IN
IA
KS
KY
ME
MA
MI
MN
MS
MO
MT
NE
NV
NH
NJ
NM
NY
NC
ND
OH
OK
OR
PA
RI
SC
TN
TX
UT
WA
WV
WI
WY
Estimated Population
80,000
Survival Method
Release Cohort Method
Figure 3: Estimated Number of Current Medium-Stay Prisoners
70,000
60,000
50,000
40,000
30,000
20,000
10,000
-
AL
CA
CO
DE
FL
GA
IN
IA
KS
KY
ME
MA
MI
MN
MS
MO
MT
NE
NV
NH
NJ
NM
NY
NC
ND
OH
OK
OR
PA
RI
SC
TN
TX
UT
WA
WV
WI
WY
Estimated Population
80,000
Survival Method
Release Cohort Method
Figure 4: Estimated Number of Current Long-Stay Prisoners
40,000
30,000
25,000
20,000
15,000
10,000
5,000
-
AL
CA
CO
DE
FL
GA
IN
IA
KS
KY
ME
MA
MI
MN
MS
MO
MT
NE
NV
NH
NJ
NM
NY
NC
ND
OH
OK
OR
PA
RI
SC
TN
TX
UT
WA
WV
WI
WY
Estimated Population
35,000
Survival Method
Release Cohort Method
| File Type | application/pdf |
| Author | carsone |
| File Modified | 2015-08-12 |
| File Created | 2015-06-26 |