documentation

Page 1 of 45
November 12, 2015
MEMORANDUM FOR THE RECORD
From:

Suzanne Dorinski
Public Sector Statistical Methods Branch
Economic Statistical Methods Division

Subject:

Documentation of the Imputation Methodology for the 2014 Juvenile Residential Facility
Census

This memorandum documents the imputation methodology for the 2014 Juvenile Residential Facility
Census (JRFC).
This document uses intentional white space to improve readability. This document is available as PDF to
preserve the original formatting.
This document also serves as a guide to the programmer who works on the 2016 JRFC. Notes to the
programmer are in brackets. The 2014 JRFC imputation system is reusable for the 2016 JRFC. [The
programs are in the \\govs009fs\smb\jrfc\2014 folder.]
The imputation methodology for the 2010 JRFC and earlier censuses used the section and question
numbers as variable names. Starting with the 2014 JRFC, variables have new names, rather than using
the section number and question number as the variable name. We needed to use prior year files for
the 2014 JRFC imputation processing, so we updated the variable names on the prior year files to use
the current naming convention. Please see the 2014 JRFC record layout for more details.
The Economic Directorate of the Census Bureau regularly conducts quality audits of survey and census
programs. Some of the discussion in this document is based on feedback from the last JRFC quality
audit. Some parts of the discussion may not be of interest to the sponsor but are required for the
purposes of the quality audit.

Page 2 of 45

Table of Contents
Introduction .................................................................................................................................................. 4
Figure 1. JRFC response method percentages over time......................................................................... 5
Figure 2. JRFC response method counts over time.................................................................................. 5
Definition of Out-of-scope Facility ............................................................................................................ 6
Unit Response Rate ................................................................................................................................... 6
Figure 3. CJRP and JRFC revised unit response rates over time............................................................... 7
Summary of Changes During 2014 Data Collection ...................................................................................... 7
Summary of the File ...................................................................................................................................... 9
Table 1. Summary of the facilities on the 2014 imputed file ................................................................... 9
Table 2. 2014 JRFC counts (nontribal facilities in the 50 states and the District of Columbia that held
juveniles on reference day)....................................................................................................................... 9
Records Eligible for Imputation .................................................................................................................... 9
Table 3. Values of the status flag for the 2014 JRFC .............................................................................. 10
Table 4. Values of the occupied flag for the 2014 JRFC ......................................................................... 10
Critical Item Facilities .................................................................................................................................. 11
Table 5. Value for CheckInStatus code................................................................................................... 11
Item Response Rates................................................................................................................................... 12
Table 6. JRFC Item Response Rates by Year ........................................................................................... 12
Figure 4. JRFC item response rates by section over time ...................................................................... 18
Questionnaire Items Eligible for Imputation .............................................................................................. 19
Table 7. Items eligible for imputation .................................................................................................... 19
Imputation Rates......................................................................................................................................... 21
Table 8. JRFC item response rates and total quantity response rates over time .................................. 22
Figure 5. JRFC key item response rates over time ................................................................................. 23
Figure 6. JRFC total quantity response rates over time ......................................................................... 23
Collapsed Facility Type Codes ..................................................................................................................... 24
Table 9. Collapsed facility type code when only one box marked ......................................................... 24
Table 10. Hierarchy used when more than one facility type marked on 2014 JRFC ............................. 24
Table 11. Frequency of methods used to assign collapsed facility code ............................................... 25
Pulling Forward Prior Year Data .................................................................................................................. 25
Table 12. Questions for which we pull forward data from prior year if current year is missing ........... 25

Page 3 of 45
Creation of Imputation Groups ................................................................................................................... 26
Table 13. Imputation variables added to the 2014 JRFC file ................................................................. 26
Collapsing of Imputation Groups ................................................................................................................ 28
Table 14. 2014 JRFC imputation groups with records needing imputation........................................... 29
Calculations within Imputation Group ........................................................................................................ 31
Table 15. Trimmed mean growth rates for prior year population counts ............................................. 32
Table 16. Selected medians ................................................................................................................... 33
Table 17. Most commonly reported answers for sleeping room arrangements ................................... 34
Table 18. Median or most commonly reported answers for large muscle activity questions .............. 35
Imputing Population Counts ....................................................................................................................... 36
Imputing the Rest of the Values for the 2014 JRFC .................................................................................... 37
Quality Checks Performed During and After Imputation............................................................................ 37
Description of the 2014 JRFC Imputed File ................................................................................................. 38
Notes on the 2014 JRFC Imputation System............................................................................................... 38
Data Quality Issues in Makeshift Beds Question ........................................................................................ 39
Recommendations for Future Collections .................................................................................................. 41
Completely revamp the facility IDs ......................................................................................................... 41
Include edit flags in the Liger export file................................................................................................. 41
Analyze the 2014 JRFC paradata ............................................................................................................. 41
Include the mathematical statistician in project meetings .................................................................... 42
Use data visualization more .................................................................................................................... 42
Consider additional edits for makeshift beds question .......................................................................... 42
Share data snapshot with facilities to help “sell” the data collections .................................................. 42
Figure 7. Number of nontribal facilities in the 50 states and District of Columbia by state juvenile
population ............................................................................................................................................... 43
Figure 8. CJRP 2013 data snapshot ........................................................................................................ 44
References .................................................................................................................................................. 45

Page 4 of 45

Introduction
First conducted in 2000, the JRFC is a mail canvass census. The 2010 JRFC was the first collection to give
facilities the option to respond online. Figure 1 shows the percentage distribution of response method
for each year since 2010. Figure 2 shows the response method counts for each year since 2010. The
online option is becoming more popular with facilities.
The JRFC asks juvenile residential custody facilities in the U.S. to describe their facilities. Adult facilities,
or facilities exclusively for drug or mental health treatment, or facilities for abused or neglected children
are not included in the census. The JRFC is conducted in even-numbered years. The reference date for
the 2014 JRFC was October 22, 2014.
The Census of Juveniles in Residential Placement (CJRP) is conducted in alternate years and describes
the characteristics of youth in custody. JRFC collects information about the facilities that hold juvenile
offenders. The 2014 JRFC asks about facility characteristics such as size, structure, type, ownership, and
security arrangements. JRFC also reports the number of juveniles who died in custody during the past
twelve months. The 2014 JRFC also includes sections that ask about mental health services, educational
services, and substance abuse services provided to the offenders.
The National Archive of Criminal Justice Data holds the previous data files, where they are part of a
restricted-access collection. For more information, see http://www.icpsr.umich.edu/icpsrweb/NACJD/.
The project sponsor is the Office of Juvenile Justice and Delinquency Prevention (OJJDP). The Public
Sector Statistical Methods Branch (PSSMB) of the Economic Statistical Methods Division is responsible
for imputation in the JRFC, while the Criminal Justice Branch (CJB) in the Economic Reimbursable
Surveys Division is responsible for the data collection and editing.
The 2014 questionnaire has seven sections:
•
•
•
•
•
•
•

Section 1, which collects general facility information,
Section 2, which collects information about mental health services,
Section 3, which collects information about educational services,
Section 4, which collects information about substance abuse services,
Section 5, which collects information about the last month,
Section 6, which collects information about the last year,
Section 7, which collects general information.

Page 5 of 45

Figure 1. JRFC response method percentages over time

Figure 2. JRFC response method counts over time

Page 6 of 45

Definition of Out-of-scope Facility
A facility is temporarily out-of-scope when they do not hold juveniles on the reference date.
A facility is permanently out-of-scope for one of several reasons:
•
•
•
•
•

The facility is no longer a residential facility (might have converted to day treatment only).
The facility is a duplicate of a record already on the data file.
The facility has changed from public to private, or private to public. When this happens, the
facility ID changes, and the previous facility ID is out-of-scope.
The facility no longer holds any juveniles (only handles adults).
The facility no longer holds any offenders (juveniles are all voluntary placements, or in the
facility because of neglect, abuse, dependency, or abandonment).

Unit Response Rate
There were 2,223 in-scope facilities on reference day. 1,942 of the 2,223 facilities responded to the
2014 JRFC, for a unit response rate of 87.4 percent. 186 of the facilities refused to participate in the
2014 JRFC. We imputed records for the nontribal facilities in the 50 states and the District of Columbia
that were refusals. There were five tribal refusal facilities and ten territorial refusal facilities. Each is on
the 2014 JRFC imputed file, but only the facility ID, contact information, method of return, the check-in
code, the status flag, and the occupied flag are available for those facilities. Other flags were set to 0 by
default to be consistent with the Liger processing system.
The unit response rate was calculated differently in previous years for both CJRP and JRFC. Previously
reported unit response rates counted critical item facilities as respondents. The revised unit response
rate is
𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 = �

(𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 + 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴)
(𝐶𝐶𝐶𝐶𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 + 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 + 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 + 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)

based on the value of the CheckInStatus code for each facility.

Figure 3 shows the revised unit response rates for CJRP and JRFC over time.
The extraordinary number of critical item facilities in the 2006 CJRP (676) explains the 76.1 percent
revised unit response rate for that year.

Page 7 of 45

Figure 3. CJRP and JRFC revised unit response rates over time
[The program JRFC_and_CJRP_unit_response_rates_over_time.sas calculates the revised unit response
rates and generates the graphic display. The program online_response_rates_over_time.sas calculates
the method of return values over time and generates the graphic displays.]

Summary of Changes During 2014 Data Collection
There were quite a few changes for the 2014 JRFC:
•
•
•
•
•

The 2014 JRFC was the first to use the Liger system for processing.
JRFC variable names are more descriptive.
The 2014 questionnaire includes three more sections.
The way we impute occupied makeshift beds has changed.
We have recalculated the CJRP and JRFC unit response rates where possible.

The 2010 JRFC and the 2012 JRFC were edited using the Harvester processing system. For 2014, the
editing was migrated to the Liger system. The flags in the Liger system were unusable. Special
processing had to be added to the imputation system to figure out which values in the file were really
missing and thus needed to be imputed. The Liger system defaults all values to 0. Without accurate
flags to indicate which values are missing, it can be difficult to distinguish between a reported value of 0
and a default value of 0 which indicates missing data. It is also worth noting that all variables on the

Page 8 of 45
Liger export file are text. Counts need to be converted to numeric for calculations within the imputation
system, and most other variables need to be converted to numeric to agree with the record layout.
The variable names for JRFC have changed completely from what they were in 2012. Since we use prior
year files while imputing the current year JRFC, we had to recode the prior year files. The 2014 JRFC
record layout includes columns for the 2013 CJRP, the 2012 JRFC, and the 2010 JRFC, with the original
variable names on those files.
The mental health services section, the educational services section, and the substance abuse services
section were not on the 2012 questionnaire. Those sections were on the 2010 questionnaire, but OJJDP
decided that the four-year gap in collection was too large to pull existing data forward for item
nonrespondents.
In the 2012 JRFC and earlier collections, we used a median value to impute occupied makeshift beds
when it was missing. The problem is that very few facilities report occupied makeshift beds, so we end
up imputing 0 values. The occupied makeshift beds question has a screener question just before it. We
ask facilities first if they had any occupied makeshift beds on reference day. If the facility answers no,
we direct them to skip over the number of occupied makeshift beds and on to the next question. The
skip pattern implies that 0 should not be a valid answer for occupied makeshift beds. See the Data
Quality Issues in Makeshift Beds Question section for more discussion.
The unit response rate reported previously for CJRP and JRFC data collections was calculated differently.
We have recalculated the rates, where possible, as discussed in the Introduction section of this
document. See Figure 3 for more details.

Page 9 of 45

Summary of the File
Table 1. Summary of the facilities on the 2014 imputed file
206
87
7
297
36
1,644
135
5
10
2
2,429

facilities permanently closed, temporarily closed, or out-of-scope
facilities that hold offenders and reported only critical items
facilities that hold nonoffenders only and reported only critical items
facilities that hold nonoffenders only and reported more than critical items
nontribal facilities in the 50 states or DC that hold nonoffenders only and were refusals
facilities that hold offenders and reported more than critical items
nontribal facilities in the 50 states or DC that hold offenders and were refusals
tribal facility refusals
territory facility refusals
facilities that either held no people or held no juveniles on reference day
facilities in the 2014 imputed file

Table 2. 2014 JRFC counts (nontribal facilities in the 50 states and the District
of Columbia that held juveniles on reference day)
63,063
333
62,730
50,821
11,909
2,191

people in residential placement
adults
juveniles
juvenile offenders
juvenile nonoffenders
nontribal facilities in the 50 states and District of Columbia that held juveniles on reference day

[The program summarize_JRFC_imputed_file.sas produces the counts in this section.]

Records Eligible for Imputation
In the previous version of JRFC, we imputed missing data for all facilities. Starting with the 2010 JRFC,
OJJDP requested that we not impute missing data for tribal facilities. The 2010 JRFC was the first cycle
to attempt to collect data from facilities in American Samoa, Guam, Northern Marianas Islands, Puerto
Rico, and the Virgin Islands. OJJDP and Census agreed that we would not impute missing data for the
territorial facilities.
Tribal facilities face special challenges; that is one reason that we do not impute them. The reader is
directed to “From Broken Homes to a Broken System”, written by Sari Horwitz and published in the
Washington Post on November 28, 2014, and accessible online at
http://www.washingtonpost.com/sf/national/2014/11/28/from-broken-homes-to-a-broken-system/.
Starting with the 2012 JRFC, we include the out-of-scope facilities, the temporarily closed facilities, and
the permanently closed facilities on the final file, even though those records are not eligible for
imputation. Researchers want to examine juvenile residential facilities over time, so we now include the
out-of-scope facilities, the temporarily closed facilities, and the permanently closed facilities so that all
facilities can be easily tracked over time.

Page 10 of 45
There are two flags on the 2014 file that help quickly classify records as eligible for imputation or not.
The status_flag_2014 describes who operates the facility. The occupied_flag_2014 indicates whether
the facility held juveniles on reference day. Table 3 explains the status flag, while Table 4 explains the
occupied flag. The first two digits of the facility ID indicate the state where the facility is located. Note
that since we do not impute missing data for tribal or territorial facilities, the occupied flag is set to 0
when the tribal or territorial facility does not answer the questions about persons assigned to beds and
juveniles assigned to beds on reference day.

Table 3. Values of the status flag for the 2014 JRFC
Status_flag_2014
Description
value
1
State-operated public facilities in the 50 states and the
District of Columbia
2
Locally-operated public facilities in the 50 states and
the District of Columbia
3
Private facilities in the 50 states and the District of
Columbia
4
Tribal facilities (missing data are not imputed)
5
Territorial facilities (missing data are not imputed)

How constructed
3rd digit of ID = 0
3rd digit of ID is 1, 2, 3, or 4
3rd digit of ID = 8
3rd digit of ID = 7
Based on state code

Table 4. Values of the occupied flag for the 2014 JRFC
Occupied_flag_2014
value
0
1

Description

How constructed

Facility had no persons assigned to beds or
had no juveniles assigned to beds on
reference day
Facility had juveniles assigned to beds on
reference day

G_ASS_BDS is missing or set to 2,
G_UNDER_21 is missing or set to 2
G_NUM_UND_21 greater than 0

Page 11 of 45

Critical Item Facilities
In follow-up interviewing, CJB attempted to collect as much data as possible to fill in the seven sections
of the questionnaire. The following data items were critical:
Section 1:
•
•
•
•
•
•
•
•
•

Question 5a and 5b (total persons assigned to beds in the facility)
Question 6 (number of persons age 21 or over assigned to beds in the facility)
Question 7a and 7b (number of persons under age 21 assigned to beds in the facility)
Questions 8a and 8b (number of offenders under age 21 assigned to beds in the facility)
Questions 9a and 9b (number of nonoffenders under age 21 assigned to beds in the facility)
Question 13 (type of facility)
Question 17a (who operates facility: private non-profit agency, for profit agency, or government
agency)
Question 22 (total number of standard beds in facility)
Question 23a and 23b (occupied makeshift beds in the facility)

Section 5:
•
•

Question 2a (any juvenile transported to hospital emergency room during September?)
Question 2b (reason why juvenile transported to hospital emergency room)

Section 6:
•
•
•

Question 1 (any juveniles die while assigned bed at facility within last 12 months?)
Question 2 (how many juveniles died while assigned bed at facility within the last 12 months)
Question 3 (cause of death, location, age, sex, race, date of admission, date of death)

The value of CheckInStatus indicates which facilities are critical item facilities. See Table 5 for more
details.

Table 5. Value for CheckInStatus code
CheckInStatus value
1
2
3
4
5
6
7
9

Meaning
Respondent
Permanently closed
Temporarily closed
Out-of-scope
Refusal
Add in (new facility) full form
Add in (new facility) critical item
Critical item

Page 12 of 45

Item Response Rates
We calculated the item response rates for the 2014 JRFC by looking at the item values before
imputation. There were 2,191 nontribal facilities in the 50 states and the District of Columbia that held
juveniles on reference day. The calculations consider the skip patterns. If a facility was skipped out of a
question on the form, they were not counted as an item nonrespondent. For example, if the facility
answered No to the question asking if they held juveniles on reference day for reasons other than
offenses (G_ASS_NONOFDRS), they were not considered an item nonrespondent for
G_NUM_ASS_NONOFDRS because they were instructed to skip G_NUM_ASS_NONOFDRS.
Table 6 shows the item response rates for both the 2012 JRFC and the 2014 JRFC. Figure 4 shows the
response rates in a graphic. Note that the item response rates were higher in 2012 except for the very
first question on the form. In Table 6, — means that question was not on the questionnaire that year.

Table 6. JRFC Item Response Rates by Year
Description

2012
2014
response response
rate
rate

S1-Question 1-Address Correct?

87.9

88.1

S1-Question 2-Physical Layout of Facility

90.4

88.3

S1-Question 3-Other bldgs w/units-beds?

90.4

88.3

S1-Question 4-Any detention overflow?

90.4

88.4

S1-Question 5a-Any persons assigned beds?

94.9

92.2

S1-Question 5b-Number assigned beds

94.8

92.2

S1-Question 6-Number assigned beds >21 yrs

94.8

92.2

S1-Question 7a-Any persons under 21 yrs?

94.8

92.2

S1-Question 7b-Number assigned <21

94.8

92.2

S1-Question 8a-Any assigned for offense?

94.8

92.1

S1-Question 8b-Number assigned for offense

94.8

92.1

S1-Question 9a-Any assigned for non-offense?

94.8

92.2

S1-Question 9b-Number assigned for non-offense

94.8

92.2

S1-Question 10a-On site residential treatment?

90.4

88.0

S1-Question10b-Type of Residential Treatment

90.4

88.0

S1-Question 11-Provide foster care?

90.4

88.0

S1-Question 12-Provide independent living?

90.4

88.0

S1-Question 13-Type of facility

94.9

91.4

S1-Question 14a-Separate housing units?

90.4

88.1

Page 13 of 45

Description

2012
2014
response response
rate
rate

S1-Question 14b-Separate housing units differ

90.2

88.1

S1-Question 14c-Purpose of separate housing

89.9

88.1

S1-Question 14d-Do separate units share

90.4

88.1

S1-Question 15a-Is facility owned by

90.4

89.8

S1-Question 15b-Name of private agency

90.4

89.8

S1-Question 16-Level of government agency

90.4

89.8

S1-Question 17a-Is facility operated by

94.9

91.0

S1-Question 17b-Name of private agency

92.6

91.0

S1-Question 18-Level of Government agency

94.2

91.0

S1-Question 19a-Any persons locked in rooms?

90.4

88.2

S1-Question 19b-When locked in rooms

90.4

88.2

S1-Question 20-Features intended to confine

90.4

88.0

S1-Question 21a-Are outside doors locked?

90.4

88.2

S1-Question 21b-Why are outside doors locked

90.4

88.2

S1-Question 21c-When are outside doors locked

90.4

88.2

S1-Question 22-Total number standard beds

94.9

92.2

S1-Question 23a-Any occupied makeshift beds

94.8

90.1

S1-Question 23b-How many makeshift beds occupied

94.8

90.1

S1-Question 24-Number of Actual Occupants per Room

90.4

88.0

S1-Question 25.-Voluntary Large Muscle Activity?

90.4

88.0

S1-Question 26a.-Required Large Muscle Activity?

90.4

88.0

S1-Question 26b-Required Minutes Per Day

90.4

88.0

S1-Question 26c.-Required Days Per Week

90.4

88.0

S2-Question 1a-Risk of Suicide

-

88.0

S2-Question 1b-What best describes the process

-

88.0

S2-Question 2-Who asks questions

-

88.0

S2-Question 3-When are young persons first asked questions

-

88.0

S2-Question 4-Which young persons are asked questions

-

88.0

S2-Question 5a-Are any young persons re-asked questions

-

88.0

S2-Question 5b-Which best describes reason for re-asking

-

88.0

Page 14 of 45

Description

2012
2014
response response
rate
rate

S2-Question 6-Does facility assign different levels of risk to young persons

-

88.0

S2-Question 7a-Are at risk young persons placed in observation room

-

88.0

S2-Question 7b-What happens in observation room

-

88.0

S2-Question 8-Preventive measures

-

87.8

S2-Question 9-Any other mental health services Inside/outside:Yes, both
available

-

88.0

S2-Question 10a-Is ongoing counseling provided

-

88.0

S2-Question 10b-Which forms of ongoing counseling

-

88.0

S2-Question 11-Are any young persons evaluated by a professional

-

88.0

S2-Question 12-When are young persons evaluated by a professional

-

87.5

S2-Question 13-Which young persons are evaluated

-

87.7

S2-Question 14a-Ongoing therapy for mental health provided

-

87.9

S2-Question 14b-Which forms of ongoing therapy are provided

-

87.9

S2-Question 14c-Which best describes the facility policy on mental health

-

87.9

S2-Question 15-Do medical professionals prescribe medication

-

88.0

S2-Question 16a-Are there one or more special living/sleeping units in the facility
reserved for mental health

-

88.0

S2-Question 16b-How do these living/sleeping units differ

-

88.0

S2-Question 17a-Is there a specialized sex offender treatment program

-

88.0

S2-Question 17b-Are any provided to young persons charged with or adjudicated
for a sex offense

-

88.0

S2-Question 18-Are there one or more special living/sleeping units for sex
offenders

-

88.0

S2-Question 19a-Are information regarding their mental health status, services
and or needs communicated to new residence or placement

-

88.0

S2-Question 19b-For which is this information shared

-

88.0

S3-Question 1-After arrival are any young persons evaluated to determine grade
level

-

87.9

S3-Question 2-When are young persons evaluated for grade level:

-

87.9

S3-Question 3-Methods of evaluation of grade level

-

87.5

S3-Question 4-Which young persons are evaluated to determine grade level

-

87.5

S3-Question 5-As part of the discharge process, are any young persons
evaluated to determine their educational grade levels and educational needs

-

87.9

Page 15 of 45

Description

2012
2014
response response
rate
rate

S3-Question 6-Which young persons are evaluated to determine grade level and
educational needs when being discharged

-

87.9

S3-Question 7a-Do any young persons receive teacher instructions at a location
either inside or outside this facility

-

87.9

S3-Question 7b-Which young persons attend school or receive teacher
instructions

-

87.9

S3-Question 8-Which of the following educational services are provided

-

87.9

S3-Question 9a.i-How many hours a week do young persons attend school or
receive teacher instruction

-

87.9

S3-Question 9a.o-How many hours a week do young persons attend school or
receive teacher instruction

-

87.9

S3-Question 9b.i-How many months per year do young persons attend school or
receive teacher instruction

-

87.9

S3-Question 9b.o-How many months per year do young persons attend school or
receive teacher instruction

-

87.9

S3-Question 10a-Upon departure, is information regarding their educational
status, services and/or needs communicated to young person's new placement

-

87.9

S3-Question 10b-For which young persons is this information shared

-

87.9

S4-Question 1a-After arrival, are any young persons evaluated to determine if
they have substance abuse problems

-

88.0

S4-Question 1b-Which methods are used to evaluate persons substance abuse
problems

-

88.0

S4-Question 2-When are young persons first evaluated to determine substance
abuse problems

-

88.0

S4-Question 3a-Are all young persons evaluated after arrival in this facility to
determine whether they have substance abuse problems

-

88.0

S4-Question 3b-Which young persons are evaluated for substance abuse
problems

-

88.0

S4-Question 4a-Are any young persons required to provide urine samples

-

88.0

S4-Question 4b-Which best describes the circumstances under which urine
samples are provided

-

88.0

S4-Question 5-Do any young persons receive substance abuse services inside or
outside this facility

-

88.0

S4-Question 6-Which of the following substance abuse services are provided

-

88.0

S4-Question 7-Which of the following self-led, self help groups are provided

-

87.9

S4-Question 8a-Is ongoing counseling for substance abuse provided

-

87.9

Page 16 of 45

Description

2012
2014
response response
rate
rate

S4-Question 8b-Which forms of ongoing counseling for substance abuse are
provided inside or outside facility

-

87.9

S4-Question 9a-Is ongoing therapy for substance abuse problems provided to
young persons

-

88.0

S4-Question 9b-Which forms of ongoing therapy for substance abuse problems
are provided

-

88.0

S4-Question 9c-Which best describes the facility policy on providing ongoing
therapy for substance abuse problems

-

88.0

S4-Question 10a-Upon departure, is information regarding their substance abuse
status, services and/or needs communicated with new placement

-

87.6

S4-Question 10b-For which young persons is this information shared

-

87.6

S5-Question 1-Any unauthorized departures?

89.9

88.1

S5-Question 2a-Any transported to hospital?

94.5

90.1

S5-Question 2b-Reason transported to hospital

94.5

90.1

S5-Question 3-Any restrained with mechanical device?

90.2

87.8

S5-Question 4-Any locked more than 4 hrs alone?

90.1

87.9

S5-Question 5a-Physical Health unavailable?

90.2

88.0

S5-Question 5b-Reasons Physical Health unavailable

90.2

88.0

S5-Question 6a-Mental Health unavailable?

90.1

87.9

S5-Question 6b-Reasons Mental Health unavailable

90.1

87.9

S5-Question 7a-Educational Instructions unavailable?

90.2

87.9

S5-Question 7b-Reasons Educational instructions unavailable

90.2

87.9

S5-Question 8a-Sub. Abuse services unavailable?

90.2

88.0

S5-Question 8b-Substance abuse services unavailable

90.2

88.0

S6-Question 1-Any persons die while assigned beds inside or outside this facility
from 10/01/2013 and 09/30/2014

94.9

90.5

S6-Question 2-How many persons died?

94.9

90.3

S6-Question 3a- Cause of death-person 1

94.9

90.3

S6-Question 3b-Location of Death of person 1

94.9

90.3

S6-Question 3C-Age at death (years) of person 1

94.9

90.3

S6-Question 3D-Sex of person 1

94.9

90.3

S6-Question 3E-Race of person 1

94.9

90.3

S6-Question 3f-Date of admission-MONTH (person 1)

94.8

90.3

Page 17 of 45

Description

2012
2014
response response
rate
rate

S6-Question 3f-Date of admission-DAY (person 1)

94.8

90.3

S6-Question 3f-Date of admission-YEAR (person 1)

94.8

90.3

S6-Question 3g-Date of Death-MONTH

94.9

90.3

S6-Question 3g-Date of Death-DAY

94.9

90.3

S6-Question 3g-Date of Death-YEAR

94.9

90.3

S6-Question 3a-Cause of Death-person 2

94.9

90.3

S6-Question 3b-Location of Death-(Person 2)

94.9

90.3

S6-Question 3c-Age at Death- (person 2)

94.9

90.3

S6-Question 3d-Sex-(person 2)

94.9

90.3

S6-Question 3e-Race-(person 2)

94.9

90.3

S6-Question 3f-Date of admission-MONTH - (person 2)

94.9

90.3

S6-Question 3f-Date of admission-DAY - (person 2)

94.9

90.3

S6-Question 3f-Date of admission-YEAR - (person 2)

94.9

90.3

S6-Question 3g-Date of Death-MONTH (person 2)

94.9

90.3

S6-Question 3g-Date of Death-DAY (person 2)

94.9

90.3

S6-Question 3g-Date of Death-YEAR (person 2)

94.9

90.3

S6-Question 3a-Cause of Death- (person 3)

94.9

90.3

S6- Question 3b-Location of Death-(person 3)

94.9

90.3

S6-Question 3c-Age at Death-(person 3)

94.9

90.3

S6-question 3d-Sex- (person 3)

94.9

90.3

S6-question 3e-Race-(person 3)

94.9

90.3

S6-Question 3f-Date of admission-MONTH- (person 3)

94.9

90.3

S6-Question 3f-Date of admission-DAY- (person 3)

94.9

90.3

S6-Question 3f-Date of admission-YEAR- (person 3)

94.9

90.3

S6-Question 3g-Date of Death-MONTH- (person 3)

94.9

90.3

S6-Question 3g-Date of Death-DAY- (person 3)

94.9

90.3

S6-Question 3g-Date of Death-YEAR- (person 3)

94.9

90.3

S7-Question 1a-Other juvenile facilities in same bldg.?

90.5

88.0

S7-Question 1b-How many other juvenile facilities

90.5

88.0

S7-Question 2-Shares the following with reported facility

90.5

88.0

Page 18 of 45

Figure 4. JRFC item response rates by section over time

Page 19 of 45

Questionnaire Items Eligible for Imputation
The following items were eligible for imputation in the 2014 JRFC. See the notes at the end of Table 7
for descriptions of the codes in the imputation method column.

Table 7. Items eligible for imputation
Question
G_PHY_LAYOUT, G_PHY_LAYOUT_SP
G_OTH_BLDGS
G_DET_OFLOW
G_NUM_ASS_BDS
G_NUM_21_OLDER

Concept
Physical layout of facility
Other buildings?
House overflow population?
Persons assigned to beds
Adults assigned to beds

G_NUM_UND_21
G_NUM_ASS_OFDRS

Juveniles assigned to beds
Juvenile offenders assigned to
beds

G_NUM_ASS_NONOFDRS

Juvenile nonoffenders
assigned to beds

G_TYP_MNTL_HLTH through G_TYP_OTH_SP
G_FOSTER_CARE
G_INDEPENDENT_LIVING
G_FAC_DETN_CTR through G_FAC_OTH_SP

On-site residential treatment?
Provide foster care?
Independent living?
Facility type

G_SEP_HOUSING through
G_SEP_N1_OF_THE_ABV
G_FAC_OWND_PRVT through
G_FAC_OWND_NAME
G_FAC_LVL_NATIVE_AMER through
G_FAC_LVL_OTH_SP
G_FAC_OPRTD_PRVT through
G_FAC_OPRTD_NAME
G_FAC_OPRTD_NATIVE_AMER through
G_FAC_OPRTD_OTH_SP
G_LCKD_RMS through G_WHEN_OTH_SP
G_CNF_SECURE_RMS through
G_CNF_N1_OF_THE_ABV
G_OUT_DRS_LCKD through
G_OUT_DRS_OTH_SP
G_STAN_BDS

Separate sleeping units?

Imputation method
A
A
A
B (flag=2)
B (flag=2)
C (flag=21)
D (flag=22)
E (flag=23)
B (flag=2)
C (flag=21)
D (flag=22)
E (flag=23)
C (flag=21)
D (flag=22)
E (flag=23)
A
A
A
See Collapsed Facility
Type Codes section
A

Facility ownership

A

Level of government that
owns facility
Who operates facility

A

Level of government that
operates facility
Juveniles locked in rooms?
Confine juveniles in specific
areas?
Outside doors locked?

A

Number of standard beds

G_MKSHFT_BDS_TOT

Occupied makeshift beds

F (flag=20)
B (flag=2)
F (flag=20)
E (flag=23)

A

A
A
A

Page 20 of 45
Question
G_NUM_1_PERSN through G_NUM_25_PLS

Concept
Sleeping room arrangements

G_LG_MUSCLE_ACT

Voluntary large muscle activity

G_REQ_LG_MUSCLE

Required large muscle activity

G_REQ_MINUTES

Minutes per day of required
large muscle activity

G_REQ_DYS_PER_WK

Days per week of required
large muscle activity

Imputation method
F (flag=20)
G (flag=24)
F (flag=20)
E (flag=23)
F (flag=20)
H (flag=30)
F (flag=20)
G (flag=24)
B (flag=2)
F (flag=20)
G (flag=24)
B (flag=2)

NOTES: A means pulled forward from prior year if missing; if no prior year available, set to refusal.
B means derived from response to other variables (flag=2)
C means imputed using growth rate applied to prior year CJRP data (flag=21)
D means imputed using growth rate applied to prior year JRFC data (flag=22)
E means imputed using median value (flag=23)
F means pulled forward data from prior year JRFC (flag=20)
G means imputed using most commonly reported answer (flag=24)
H means randomly assigned based on reported rates (flag=30)

Page 21 of 45

Imputation Rates
Table 8 shows the item response rates and the total quantity response rates for 2012 and 2014.
The item response rate is
𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁 𝑜𝑜𝑜𝑜 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑤𝑤𝑤𝑤𝑤𝑤ℎ 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑓𝑓𝑓𝑓𝑟𝑟 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖
× 100.
𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁 𝑜𝑜𝑜𝑜 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 𝑓𝑓𝑓𝑓𝑓𝑓 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝐽𝐽𝐽𝐽𝐽𝐽𝐽𝐽 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓

The item imputation rate is 100 – the item response rate. For the 2014 JRFC, the item imputation rate is
7.8 percent for all key items except for makeshift beds, which has an item imputation rate of 9.9
percent.
The total quantity response rate shows how much of the quantity estimate was reported. The total
quantity response rate is
𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 𝑑𝑑𝑑𝑑𝑑𝑑𝑎𝑎 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑏𝑏𝑏𝑏 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓
× 100.
𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 𝑏𝑏𝑏𝑏𝑏𝑏ℎ 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑎𝑎𝑎𝑎𝑎𝑎 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑

Table 8 shows that we did not impute any adults assigned to beds. This is not surprising, given how few
facilities reported adults assigned to beds in the 2014 JRFC.
Figure 5 shows that the key item response rates were higher in 2012 than 2014. Figure 6 shows the
same pattern for the total quantity response rate.
[The program JRFC_response_rates_over_time.sas calculates the item response rates and the total
quantity response rates and produces the graphic displays.]

Page 22 of 45

Table 8. JRFC item response rates and total quantity response rates over time
2012 JRFC

Variable

Total
Quantity
Item
Reported Imputed
Total Response Reported Imputed Quantity Response
Records Records Facilities
Rate
Count
Count Estimate
Rate

S1-Question 5b-Number
assigned beds

2,216

121

2,337

94.8

67,253

2,559

69,812

96.3

S1-Question 6-Number
assigned beds >21 yrs

2,216

121

2,337

94.8

196

0

196

100.0

S1-Question 7b-Number
assigned <21

2,215

122

2,337

94.8

67,001

2,615

69,616

96.2

S1-Question 8b-Number
assigned for offense

2,216

121

2,337

94.8

55,747

1,443

57,190

97.5

S1-Question 9b-Number
assigned for non-offense

2,216

121

2,337

94.8

11,310

1,116

12,426

91.0

S1-Question 22-Total
number standard beds

2,217

120

2,337

94.9

97,168

3,331

100,499

96.7

S1-Question 23b-How
many makeshift beds
occupied

2,216

121

2,337

94.8

82

0

82

100.0

2014 JRFC

Variable

Total
Item
Quantity
Reported Imputed
Total Response Reported Imputed Quantity Response
Records Records Facilities
Rate
Count
Count Estimate
Rate

S1-Question 5b-Number
assigned beds

2,020

171

2,191

92.2

59,281

3,782

63,063

94.0

S1-Question 6-Number
assigned beds >21 yrs

2,020

171

2,191

92.2

333

0

333

100.0

S1-Question 7b-Number
assigned <21

2,020

171

2,191

92.2

58,948

3,782

62,730

94.0

S1-Question 8b-Number
assigned for offense

2,020

171

2,191

92.2

48,324

2,497

50,821

95.1

S1-Question 9b-Number
assigned for non-offense

2,020

171

2,191

92.2

10,624

1,285

11,909

89.2

S1-Question 22-Total
number standard beds

2,020

171

2,191

92.2

88,676

4,772

93,448

94.9

S1-Question 23b-How
many makeshift beds
occupied

1,974

217

2,191

90.1

82

65

147

55.8

Page 23 of 45

Figure 5. JRFC key item response rates over time

Figure 6. JRFC total quantity response rates over time

Page 24 of 45

Collapsed Facility Type Codes
For imputation purposes, we assign a collapsed facility type code (COLLAPSED_FACILITY) to every
facility. We only assign COLLAPSED_FACILITY to nontribal facilities in the 50 states and the District of
Columbia in the 2014 JRFC data file, because we did not impute the tribal facilities or the territorial
facilities. COLLAPSED_FACILITY is the same as the cat code on the CJRP data files.
The following procedure assigns the COLLAPSED_FACILITY code:
1. If the agency checks only one facility type box on the 2014 form, we map the checked box to the
appropriate collapsed facility type and COLLAPSED_FACILITY is set. See Table 9. Note that if the
agency only checks “other”, we do not assign a collapsed facility type here.

Table 9. Collapsed facility type code when only one box marked
COLLAPSED_FACILITY
Collapsed Facility Type
0
Detention Center
1
Shelter
2
3
5
6

Reception / Diagnostic
Center
Training School
Ranch, Camp, or Farm
Halfway House / Group
Home

Check box on 2012 form
G_FAC_DETN_CTR
G_FAC_RUNAWAY,
G_FAC_OTH_SHELTER
G_FAC_RCPTN
G_FAC_TRNING_SCH
G_FAC_BTCMP, G_FAC_RANCH
G_FAC_GP_HOME, G_FAC_RES_TRMT

2. If the agency checks more than one facility type box on the 2014 form, we determine the
collapsed facility type based on a hierarchy. The collapsed facility type code is determined by
the highest checked box on the list. If an agency checks boxes that indicate that it is both a
reception center and a training school, the assigned code is training school, since the training
school is higher up on the hierarchy than reception center is. Table 10 shows the hierarchy.

Table 10. Hierarchy used when more than one facility type marked on
2014 JRFC
COLLAPSED_FACILITY
Collapsed Facility Type
3
Training School
0
Detention Center
2
Reception / Diagnostic
Center
5
Ranch, Camp, or Farm
1
Shelter
6

Halfway House / Group
Home

Check box on 2012 form
G_FAC_TRNING_SCH
G_FAC_DETN_CTR
G_FAC_RCPTN
G_FAC_BTCMP, G_FAC_RANCH
G_FAC_RUNAWAY,
G_FAC_OTH_SHELTER
G_FAC_GP_HOME, G_FAC_RES_TRMT

Page 25 of 45

3. If the collapsed facility type is still missing, we use the facility’s cat code from the 2013 CJRP.
The facility’s answers to the facility type question from the 2013 CJRP are pulled forward.
4. If the collapsed facility type is still missing, we base the COLLAPSED_FACILITY on the facility’s
answers in the 2012 JRFC. The facility’s answers in the 2012 JRFC are pulled forward.
5. If the collapsed facility type is still missing, OJJDP will assign the code. (This situation occurred
for one facility in the 2014 JRFC.)
Table 11 shows the frequency of the methods used to assign collapsed facility type in the 2014 JRFC.

Table 11. Frequency of methods used to assign collapsed facility code
Method of assigning collapsed facility type
Facility checked one facility type box on 2014 JRFC form
Facility checked multiple facility type boxes on 2014 JRFC form
Facility responses from the 2013 CJRP were used
Facility responses from the 2012 JRFC were used
Collapsed facility type based on guidance from OJJDP
Total

Frequency

1,652
321
212
5
1
2,191

[The program assign_collapsed_facility_type_if_missing.sas creates the COLLAPSED_FACILITY variable.]

Pulling Forward Prior Year Data
The imputation process starts by pulling forward prior year data for refusals and critical item facilities
(CheckInStatus codes of 5, 7, and 9) and for item nonrespondents.

Table 12. Questions for which we pull forward data from prior year if current
year is missing
Question
G_PHY_LAYOUT, G_PHY_LAYOUT_SP
G_OTH_BLDGS
G_DET_OFLOW
G_ON_SITE_RES_TRTMNT through G_TYP_OTH_SP
G_FOSTER_CARE
G_INDEPENDENT_LIVING
G_SEP_HOUSING through G_SEP_N1_OF_THE_ABV
G_FAC_OWND_PRVT through G_FAC_OWND_NAME
G_FAC_LVL_NATIVE_AMER through G_FAC_LVL_OTH_SP
G_FAC_OPRTD_PRVT through G_FAC_OPRTD_NAME
G_FAC_OPRTD_NATIVE_AMER through
G_FAC_OPRTD_OTH_SP
G_LCKD_RMS through G_WHEN_OTH_SP
G_CNF_SECURE_RMS through G_CNF_N1_OF_THE_ABV
G_OUT_DRS_LCKD through G_OUT_DRS_OTH_SP

Concept
Physical layout of facility
Other buildings?
House overflow population?
On-site residential treatment?
Provide foster care?
Independent living?
Separate sleeping units?
Facility ownership
Level of government that owns facility
Who operates facility
Level of government that operates facility
Juveniles locked in rooms?
Confine juveniles in specific areas?
Outside doors locked?

Page 26 of 45

If the same question appears on the 2013 CJRP, we check the facility’s answers on that first, before
checking the 2012 JRFC. If the question does not appear on the 2013 CJRP, we check the facility’s
answers on the 2012 JRFC. Note that a facility may report some of the items in the current year, so the
program has to check to see if the item is missing in the current year.
If no prior year data exists for the question, we set the value to 8 (refusal). The program also verifies
that we did not overwrite any reported data for those questions.
[The program pull_forward_general_facility_info.sas does the imputations for the questions listed in
Table 12.]

Creation of Imputation Groups
The next step in the process is to create imputation groups. We create a two-digit state code from the
facility ID (note that these are not ANSI FIPS codes). We use the two-digit state code to assign a division
code to each facility. We then use the division code to assign a region code to each facility. Each facility
receives an imputation group code (IMPGRP), which is 4 characters long. The first character of IMPGRP
is the region code, the second character is the division code, and the third and fourth characters are the
collapsed facility code, formatted with a leading 0. Table 13 summarizes the imputation variables added
to the 2014 JRFC file.

Table 13. Imputation variables added to the 2014 JRFC file
Variable
COLLAPSED_FACILITY

IMPSET

Description
We collapse the facility type answers, following the CJRP methodology. The
values are:
0 – Detention Center
1 – Shelter
2 – Reception / Diagnostic Center
3 – Training School
5 – Ranch, Camp, or Farm
6 – Halfway House/ Group Home
Indicates if the record needed imputation. Value is “needs” when record needs
imputation, “donor” when record was used in growth rate calculations. Value is
blank for records that were not eligible for imputation.

Page 27 of 45
Variable
DIVISION

REGION

FACILITY_CODE
IMPGRP

Description
Census division code. The values are:
1 – New England
2 – Middle Atlantic
3 – East North Central
4 – West North Central
5 – South Atlantic
6 – East South Central
7 – West South Central
8 – Mountain
9 – Pacific
See http://www.census.gov/geo/maps-data/maps/docs/reg_div.txt for more
details.
Census region code. The values are:
1 – Northeast
2 – Midwest
3 – South
4 – West
See http://www.census.gov/geo/maps-data/maps/docs/reg_div.txt for more
details.
Collapsed facility code, formatted with a leading 0.
Imputation group code. The imputation group code is 4 characters. The first
character is the region code, the second character is the division code, and third
and fourth characters are FACILITY_CODE. If we need to collapse the
imputation group, we collapse first within the region, and the division code is
set to 0. If we need to collapse further, we collapse to the national level, and
the region code is set to 0.

[The program impute_JRFC_counts.sas creates the variables listed in Table 13.]

Page 28 of 45

Collapsing of Imputation Groups
As mentioned in Table 13, we may need to collapse the imputation group. We require an imputation
group to have at least 15 respondents and at least a 75 percent response rate within the group. If the
group does not meet either criterion, we collapse the imputation group until it meets both criteria. We
first collapse within the region, and set the division code to 0. If we need to collapse further, we
collapse to the national level, and the region code is set to 0.
The impset variable divides the records eligible for imputation into two groups: donor records and
records that need imputation. We classify a facility as needs imputation if either of the following
applies:
•
•

Facility is a refusal or critical item facility (CheckInStatus of 5, 7, or 9)
Any of the following counts are missing:
o Number of persons assigned to beds (G_NUM_ASS_BDS)
o Adults assigned to beds (G_NUM_21_OLDER)
o Number of juveniles assigned to beds (G_NUM_UND_21)
o Number of juvenile offenders assigned to beds (G_NUM_ASS_OFDRS)
o Number of juveniles assigned to beds for reasons other than offenses
(G_NUM_ASS_NONOFDRS)
o Number of standard beds (G_STAN_BDS).

In 2012, if the number of occupied makeshift beds was missing, we classified the facility as needs
imputation. That proved unworkable in 2014 when we made a late decision to change how we impute
occupied makeshift beds.
Table 14 shows the imputation groups for the 2014 JRFC with records that need imputation.

Page 29 of 45

Table 14. 2014 JRFC imputation groups with records needing imputation
Imputation
Group
Description

Records
needing
Percent
Donors imputation responding

0002

REGION= Nationwide
DIVISION=
Collapsed facility= Reception / Diagnostic Center

33

5

86.8

0005

REGION= Nationwide
DIVISION=
Collapsed facility= Ranch, Camp, or Farm

31

2

93.9

1003

REGION= Northeast
DIVISION=
Collapsed facility= Training School

25

1

96.2

1006

REGION= Northeast
DIVISION=
Collapsed facility= Halfway House / Group Home

211

34

86.1

1200

REGION= Northeast
DIVISION= Middle Atlantic
Collapsed facility= Detention Center

45

8

84.9

2300

REGION= Midwest
DIVISION= East North Central
Collapsed facility= Detention Center

102

8

92.7

2306

REGION= Midwest
DIVISION= East North Central
Collapsed facility= Halfway House / Group Home

133

6

95.7

2400

REGION= Midwest
DIVISION= West North Central
Collapsed facility= Detention Center

53

1

98.1

2401

REGION= Midwest
DIVISION= West North Central
Collapsed facility= Shelter

23

2

92.0

2403

REGION= Midwest
DIVISION= West North Central
Collapsed facility= Training School

16

1

94.1

2406

REGION= Midwest
DIVISION= West North Central
Collapsed facility= Halfway House / Group Home

149

23

86.6

3003

REGION= South
DIVISION=
Collapsed facility= Training School

39

9

81.3

3006

REGION= South
DIVISION=
Collapsed facility= Halfway House / Group Home

239

68

77.9

Page 30 of 45

Imputation
Group
Description

Records
needing
Percent
Donors imputation responding

3500

REGION= South
DIVISION= South Atlantic
Collapsed facility= Detention Center

80

18

81.6

3501

REGION= South
DIVISION= South Atlantic
Collapsed facility= Shelter

30

4

88.2

3600

REGION= South
DIVISION= East South Central
Collapsed facility= Detention Center

48

1

98.0

3700

REGION= South
DIVISION= West South Central
Collapsed facility= Detention Center

80

5

94.1

3701

REGION= South
DIVISION= West South Central
Collapsed facility= Shelter

29

2

93.5

4001

REGION= West
DIVISION=
Collapsed facility= Shelter

31

5

86.1

4006

REGION= West
DIVISION=
Collapsed facility= Halfway House / Group Home

217

54

80.1

4800

REGION= West
DIVISION= Mountain
Collapsed facility= Detention Center

56

2

96.6

4803

REGION= West
DIVISION= Mountain
Collapsed facility= Training School

26

2

92.9

4900

REGION= West
DIVISION= Pacific
Collapsed facility= Detention Center

94

4

95.9

265

[We collapse the imputation groups in the impute_JRFC_counts.sas program.]

Page 31 of 45

Calculations within Imputation Group
Once we have formed imputation groups, we calculate the following growth rates for the donors in each
imputation group:
•
•
•

Adults assigned to beds (G_NUM_21_OLDER)
Number of juvenile offenders assigned to beds (G_NUM_ASS_OFDRS)
Number of juveniles assigned beds for reasons other than offenses (G_NUM_ASS_NONOFDRS).

Since these counts are on both the CJRP and JRFC prior year files, we can calculate growth rates from
either prior year file. We use PROC UNIVARIATE in SAS to calculate trimmed means within each
imputation group, where the trimmed mean removes two observations from each imputation group.
We calculate the ratio of standard beds (G_STAN_BDS) to people assigned to beds (G_NUM_ASS_BDS)
for each donor facility in each imputation group.
We calculate the following median values for donors in each imputation group:
•
•

Occupied makeshift beds (G_MKSHFT_BDS_TOT)
Voluntary large muscle activity (G_LG_MUSCLE_ACT).

We calculate the following most commonly reported answers for donors in each imputation group:
•
•
•

Sleeping room arrangements (G_NUM_1_PERSN through G_NUM_25_PLS)
Minutes per day of required large muscle activity (G_REQ_MINUTES)
Days per week of required large muscle activity (G_REQ_DYS_PER_WK).

We calculate the required exercise rate for donor records within each imputation group. The required
exercise rate is the number of facilities that answered Yes to G_REQ_LG_MUSCLE divided by the number
of facilities that answered G_REQ_LG_MUSCLE.
Table 15, Table 16, Table 17, and Table 18 show the calculations by imputation group for the 2014 JRFC.

Page 32 of 45

Table 15. Trimmed mean growth rates for prior year population counts
Growth rate from prior CJRP

IMPGRP

G
NUM
21
OLDER

G
NUM
ASS
OFDRS

G
NUM
ASS
NONOFDRS

Growth rate from prior JRFC
G
NUM
21
OLDER

G
NUM
ASS
OFDRS

G
NUM
ASS
NONOFDRS

0002

1.000000

0.993168

0.953631

1.000000

0.955367

1.006469

0005

1.000000

0.916414

1.014819

1.000000

0.893702

1.000000

1003

0.900000

0.929036

1.000000

1.000000

0.965154

1.000000

1006

0.978723

0.953252

1.007896

1.000000

0.971632

1.007565

1200

1.000000

0.942167

1.000000

1.000000

1.018994

0.937500

2300

0.989247

1.095819

1.000614

1.000000

1.063534

0.985056

2306

1.000000

0.935009

1.137041

0.992000

1.061248

1.030759

2400

1.000000

1.133436

0.989409

1.000000

1.107901

0.900650

2401

1.000000

0.749940

1.160998

1.000000

0.952216

0.822928

2403

1.000000

0.953486

1.000000

1.000000

0.942807

1.000000

2406

0.987654

0.878601

1.011482

1.000000

0.935287

0.989132

3003

1.000000

0.989297

1.000000

1.000000

1.013790

1.000000

3006

1.000000

0.969893

0.962326

1.000000

0.921397

0.994334

3500

1.000000

1.070576

1.000000

1.000000

1.070129

1.000000

3501

1.000000

0.931696

1.242808

1.000000

0.921948

0.866587

3600

1.000000

1.102432

1.000000

1.000000

0.969256

1.000000

3700

1.000000

1.111166

1.000000

1.000000

1.051505

0.931507

3701

1.000000

0.954778

0.884516

1.000000

0.952028

1.127234

4001

1.000000

0.738127

1.607442

1.000000

0.864514

1.060603

4006

0.995968

1.013641

1.133325

1.000000

0.888080

1.009357

4800

1.000000

1.035419

1.000000

1.000000

1.113603

1.000000

4803

1.000000

0.907848

1.000000

1.000000

0.842739

1.000000

4900

1.000000

1.069448

0.978927

1.000000

0.998697

0.926610

Page 33 of 45

Table 16. Selected medians
Medians based on 2014 JRFC

IMPGRP

G
NUM
21
OLDER

G
NUM
ASS
OFDRS

(G
STAN
BDS /
G
NUM
ASS
BDS )

G
NUM
ASS
NONOFDRS

G
MKSHFT
BDS
TOT

0002

0

8.0

0

1.300000

0005

0

30.0

0

1.333333

1003

0

21.0

0

1.631579

1006

0

7.0

3

1.230769

1200

0

12.0

0

1.923077

2300

0

23.0

0

1.411765

3

2306

0

6.0

5

1.261905

2

2400

0

13.0

0

1.750000

1.5

2401

0

1.0

5

1.500000

2403

0

42.0

0

1.348485

2406

0

8.0

4

1.200000

3003

0

53.0

0

1.181818

3006

0

8.0

0

1.142857

2

3500

0

27.0

0

1.500000

4

3501

0

4.0

3

1.215385

1

3600

0

14.5

0

2.341667

2

3700

0

20.0

0

1.381628

8

3701

0

0.0

5

1.666667

4001

0

0.0

6

1.588235

4006

0

5.0

2

1.200000

4800

0

12.5

0

1.643750

9

4803

0

51.5

0

1.390097

4

4900

0

23.5

0

2.027397

1

4

1

Page 34 of 45

Table 17. Most commonly reported answers for sleeping room arrangements
Most common G NUM

IMPGRP
0002

1
PERSN

2
PERSN

3
PERSN

5
10
PERSN

11
25
PERSN

25
PLS

1

0005

1

1003

1

1006

1

1200

1

2300

1

2306

1

2400

1

2401

1

2403

1

2406

1

3003

1

3006
3500

4
PERSN

1

1

1
1

3501

1

3600

1

3700

1

3701

1

4001

1

1

4006

1

1

4800

1

4803

1

4900

1

Page 35 of 45

Table 18. Median or most commonly reported answers for large muscle
activity questions
Most common

IMPGRP

G
LG
MUSCLE
ACT
median

Required
exercise
rate

G
REQ
DYS
PER
WK

G
REQ
MINUTES

0002

1

0.515152

60

7

0005

1

0.645161

60

7

1003

1

0.560000

60

7

1006

1

0.383886

60

7

1200

1

0.533333

60

7

2300

1

0.784314

60

7

2306

1

0.338346

60

7

2400

1

0.566038

60

7

2401

1

0.217391

60

7

2403

1

0.687500

60

7

2406

1

0.523490

60

7

3003

1

0.743590

60

7

3006

1

0.476987

60

7

3500

1

0.625000

60

7

3501

1

0.200000

30

7

3600

1

0.604167

60

7

3700

1

0.675000

60

7

3701

1

0.344828

60

7

4001

1

0.193548

60

7

4006

1

0.382488

60

7

4800

1

0.732143

60

7

4803

1

0.846154

60

7

4900

1

0.702128

60

7

Page 36 of 45

Imputing Population Counts
Once we have finished the calculations within imputation group, we can impute the population counts
on the file. We start by deriving counts when facilities report only some of the counts. If the number of
juvenile offenders (G_NUM_ASS_OFDRS) is greater than 0 and the number of juveniles assigned beds for
reasons other than offenses (G_NUM_ASS_NONOFDRS) is greater than 0 but the number of juveniles
assigned to beds (G_NUM_UND_21) is not reported, we add G_NUM_ASS_OFDRS and
G_NUM_ASS_NONOFDRS to derive G_NUM_UND_21 and set the flag for G_NUM_UND_21 to derived.
If the number of persons assigned to beds (G_NUM_ASS_BDS) is the same as the number of juveniles
assigned to beds (G_NUM_UND_21) but adults assigned to beds (G_NUM_21_OLDER) is missing, we set
G_NUM_21_OLDER to 0 and set the flag to derived.
If adults assigned to beds (G_NUM_21_OLDER) is still missing, we impute a value based on one of the
following methods:
•
•
•

If prior year CJRP data are available, we apply the CJRP G_NUM_21_OLDER growth rate to the
prior year value.
If no prior year CJRP data are available but prior year JRFC data are available, we apply the JRFC
G_NUM_21_OLDER growth rate to the prior year value.
If neither prior year CJRP nor prior year JRFC data are available, we use the rounded
G_NUM_21_OLDER median value.

If the number of juvenile offenders (G_NUM_ASS_OFDRS) is still missing, we impute a value based on
one of the following methods:
•
•
•

If prior year CJRP data are available, we apply the CJRP G_NUM_ASS_OFDRS growth rate to the
prior year value.
If no prior year CJRP data are available, but prior year JRFC data are available, we apply the JRFC
G_NUM_ASS_OFDRS growth rate to the prior year value.
If neither prior year CJRP nor prior year JRFC data are available, we use the rounded
G_NUM_ASS_OFDRS median value.

If the number of juveniles assigned to beds for reasons other than offenses (G_NUM_ASS_NONOFDRS)
is still missing, we impute a value based on one of the following methods:
•
•
•

If prior year CJRP data are available, we apply the CJRP G_NUM_ASS_NONOFDRS growth rate to
the prior year value.
If no prior year CJRP data are available, but prior year JRFC data are available, we apply the JRFC
G_NUM_ASS_NONOFDRS growth rate to the prior year value.
If neither prior year CJRP data nor prior year JRFC data are available, we use the rounded
G_NUM_ASS_NONOFDRS median value.

If the number of juveniles assigned to beds (G_NUM_UND_21) is still missing, we add
G_NUM_ASS_OFDRS and G_NUM_ASS_NONOFDRS and set the flag for G_NUM_UND_21 to derived.

Page 37 of 45
If the number of persons assigned to beds (G_NUM_ASS_BDS) is still missing, we add G_NUM_UND_21
and G_NUM_21_OLDER and set the flag for G_NUM_ASS_BDS to derived.
If the number of total standard beds (G_STAN_BDS) is still missing, we pull forward the prior year JRFC
data if available; otherwise we apply the median beds to people ratio to the number of persons assigned
to beds (G_NUM_ASS_BDS) and set the flag for G_STAN_BDS appropriately.
If the number of occupied makeshift beds (G_MKSHFT_BDS_TOT) is still missing, we pull forward the
prior year JRFC data if available. If no prior year data are available from JRFC, we use the median value
for G_MKSHFT_BDS_TOT and set the flag for G_MKSHFT_BDS_TOT appropriately.

Imputing the Rest of the Values for the 2014 JRFC
If the answer to the question about young persons having opportunities for voluntary participation in
large muscle activity (G_LG_MUSCLE_ACT) is missing, we pull forward the answer from the prior year
JRFC if available. If the prior year JRFC data are unavailable, we use the median value within the
imputation group.
If the answers to the sleeping room arrangements question (G_NUM_1_PERSN through
G_NUM_25_PLS) are missing, we pull forward the prior year JRFC data if available. If no prior year JRFC
data are available, we use the most commonly reported answer in the imputation group and set
G_FLG_NUM_PERSN appropriately.
If the answers to the required participation in large muscle activity questions (G_REQ_LG_MUSCLE,
G_REQ_MINUTES, G_REQ_DYS_PER_WK) are missing, we pull forward the data from the prior year JRFC.
If no prior year data are available, we randomly assign an answer to G_REQ_LG_MUSCLE based on the
required_exercise_rate in the imputation group. If the answer to G_REQ_LG_MUSCLE is Yes and the
answers to G_REQ_MINUTES and G_REQ_DYS_PER_WK are still missing, we use the most commonly
reported answers for G_REQ_MINUTES and G_REQ_DYS_PER_WK and set the flags appropriately. If the
answer to G_REQ_LG_MUSCLE is No and the answers to G_REQ_MINUTES and G_REQ_DYS_PER_WK
are still missing, we set both G_REQ_MINUTES and G_REQ_DYS_PER_WK to 0 and set the flags for
G_REQ_MINUTES and G_REQ_DYS_PER_WK to derived.
[The program impute_JRFC_counts.sas does the calculations within imputation groups and then
completes the imputations.]

Quality Checks Performed During and After Imputation
The program that assigns COLLAPSED_FACILITY checks that all records eligible for imputation have
collapsed facility type assigned. If there are records for which collapsed facility type is missing, we
generate a report for analyst review.
The program verify_JRFC_imputations_complete.sas runs after imputation of the missing values, to
check that all the flags are properly set and that all imputed fields have valid values.

Page 38 of 45
The program check_2014_JRFC_logical_edits_AFTER_IMPUTATION.sas performs a variety of checks on
the imputed data. It looks for mismatches between the screener questions and the follow-up questions.
It checks the facility counts: total persons assigned to beds is the sum of adults and juveniles, the sum of
juvenile offenders and juvenile nonoffenders sums to the number of juveniles.

Description of the 2014 JRFC Imputed File
For ease of use, OJJDP requested an Excel spreadsheet for the 2014 JRFC imputed file, in addition to the
ASCII file. Note that the Excel spreadsheet is sorted by facility ID. Use the status_flag_2014 and
occupied_flag_2014 to group the records for analysis.

Notes on the 2014 JRFC Imputation System
We used SAS 9.4M2 in the Census Bureau’s Virtual Desktop Infrastructure (VDI) environment to do the
imputation processing.
We edited a copy of the record layout provided by CJB so that the imputation system could read it
directly into SAS. Using the metadata directly from the modified record layout allows us to minimize
hardcoding values in the SAS programs. We used the record layout metadata to:
•
•
•
•
•
•

put the variables in the proper record layout order,
convert text fields to numeric,
recode items where 0 is not a valid answer,
recode variables on the prior year files to match the new variable naming conventions,
provide explanatory text in the table of response rates by item,
dynamically create the put statement used to create the ASCII final imputed file.

See Dilorio and Abolafia (2004) and Dilorio and Abolafia (2006) for good background papers on using
metadata in SAS programming. Other examples by this author are online at
http://www.sascommunity.org/wiki/Presentations:Dorinski_Papers_and_Presentations.
The imputation system consists of a driver program written in SAS that calls all the other programs in
turn. We redirect the SAS log to a file, which we automatically convert into PDF. We use a date stamp
on the output files, so that we can rerun the system if needed without overwriting earlier results.
The 2014 imputation system can run as quickly as 5 minutes. In the author’s experience, the VDI servers
tend to be busiest from 10 am to noon and from 1 pm to 3 pm. When running the imputation system
during those hours, it can take as long as 13 minutes. The load on the VDI servers is lighter as the
afternoon progresses and employees with earlier tours of duty finish their days. The author has found
quicker run times when starting the imputation system after 4:30 pm. [The next version of the VDI
environment should result in more consistent run times. Ten to twelve employees currently share a
server, and one employee performing a resource-intensive task can affect the other employees on the
server. In the next version of VDI, each employee on the server will have a resource allocation limit, so
that one employee is not slowing down the other employees on the shared server.]

Page 39 of 45
We use the ODS Excel destination and the ODS PDF destination to produce reports for analyst review.
The imputation processing is an iterative process, with the analysts researching unusual values in the
edited file before the production imputation system run. We generate RTF files for the tables that are
included in this document. We generate PNG files using PROC SGPLOT for the graphics in this
document. We use PROC EXPORT to create the Excel spreadsheet version of the final imputed file. We
use PROC REPORT and the PDF destination to produce the more than 150 tables that accompany the
final imputed file.

Data Quality Issues in Makeshift Beds Question
According to the 2010 JRFC imputation documentation, the value for occupied makeshift beds was
imputed for 21 facilities by deriving the value. The SAS programs for the 2010 JRFC imputation
processing show that the refusal facilities were assigned an answer of No for the screener question,
unless the number of persons assigned to beds was greater than the standard number of beds, in which
case occupied makeshift beds was set to the difference between persons assigned to beds and standard
beds.
For the 2012 JRFC, the author noticed that very few facilities were reporting occupied makeshift beds,
so she asked Heather West 1 for guidance on how to impute occupied makeshift beds if the facility did
not answer the screener question. Heather mentioned that the number of standard beds is similar to
capacity in jails, and would not change unless the facility added a wing or closed a section. We imputed
makeshift beds in 2012 by pulling forward prior year data if available, or using the median value in the
imputation cell if prior year data were not available. The result was that no facility was imputed as
having occupied makeshift beds.
For the 2014 JRFC, the author noticed that the total number of reported occupied makeshift beds
remained the same, and under the prior imputation methodology, no facility was imputed as having
occupied makeshift beds. That prompted the author to look more closely at the question wording, and
then at the record layouts.
The 2012 record layout allowed 0 as a valid value for the occupied makeshift beds field. The 2014
record layout allowed 0 as a valid value for the occupied makeshift beds field. With so few facilities
reporting occupied makeshift beds, the calculation of the median within imputation groups is sensitive
to whether or not the 0 values are included in the calculation. If 0 is a valid value for occupied makeshift
beds and the screener question is ignored, every imputation group has a median value of 0 occupied
makeshift beds. Table 16 shows the median occupied makeshift beds within imputation group when the
0 values are excluded from the calculations.
As the author pondered which methodology would be better, she searched online for any mentions of
occupied makeshift beds from the JRFC collections. The search yielded two documents by Joseph
Moone: a Fact Sheet published by OJJDP in 2000, and a paper delivered at the International Conference
on Establishment Surveys (ICES), also in 2000. The Fact Sheet mentioned that the JRFC included
1

Heather West worked at the Bureau of Justice Statistics before joining Census.

Page 40 of 45
information on the use of bedspace in the facility as an indication of whether the facility is experiencing
crowding.
As the ICES paper describes, the occupied makeshift beds question was a compromise when trying to
measure the capacity of the juvenile residential facility. Research in the late 1990s showed that juvenile
residential facilities have difficulty reporting their capacity. Research also revealed the difficulty in
defining a statistical reporting unit for a juvenile residential facility.
The Census Bureau served as the data collection agent for the Annual Survey of Jails (ASJ) from the first
data collection in 1982 through the 2014 data collection cycle. The Bureau of Justice Statistics is the
sponsor of the ASJ collection. Dorinski worked on the ASJ data collections from the 2000 collection
through the 2014 collection. The edit on the capacity question in the ASJ asked jails to confirm any
change in capacity from year to year, even if the increase or decrease was as small as one inmate.
The JRFC edits do not compare the standard beds and occupied makeshift beds data to the prior year
JRFC data. The JRFC edits do not prompt facilities to explain why there are occupied makeshift beds in
use if the number of persons assigned to beds on reference day is less than the number of standard
beds. Facilities may have good reasons for having occupied makeshift beds when some of the standard
beds are not in use. As the ICES paper points out, some facilities perform different functions for
different residents. It is possible that the standard beds are assigned by function, and some areas of the
facility may be crowded (using occupied makeshift beds because all the standard beds for that function
are in use), while other areas have standard beds not in use on reference day. However, it is also
possible that one of the counts is incorrect. We recommend an edit to confirm reported occupied
makeshift beds when the number of standard beds is greater than the number of persons assigned to
beds on reference day. We also recommend edits that compare the data to prior year JRFC data.
There has been substantial turnover in project staff at Census since the beginning of the CJRP data
collection in 1997 and the beginning of the JRFC data collection in 2000. The importance of the
makeshift beds question as a compromise measure in place of facility capacity may have been
overlooked over time. The Census Bureau has a SharePoint team site for the Juvenile Frame
Development project. The site contains a library for all the CJRP and JRFC imputation methodology
documentation. The evaluation reports produced by the Census Bureau from the original CJRP
questionnaire design are available on the team site, along with the two documents by Joseph Moone.

Page 41 of 45

Recommendations for Future Collections
Completely revamp the facility IDs
The fact that a change in ownership from private to public or vice versa causes the ID to change is
problematic. Historically, CJRP and JRFC have been treated as completely separate data collections that
just happen to target the same facilities. It is clear that researchers are interested in analyzing facilities
longitudinally. That is difficult to do when the facility ID changes. The facility ID uses a two-digit state
code created by Governments Division. That code has been confused with the Federal Information
Processing Series (FIPS) state code, also a two-digit code 2.
The only information that should be encoded in a facility ID is the state where the facility is located.
Encoding any other information (e.g. type of facility, or an indicator that the facility serves juveniles) will
simply recreate the problems with IDs that change over time.

Include edit flags in the Liger export file
Currently there is no way to quantify how much of the data in the JRFC file has been edited, because the
flags in the record layout that were included in the Liger processing system do not keep track of the
editing done by the analysts. The 2014 JRFC flags from the processing system were unusable. The
imputation system had to have special routines to figure out what data was actually missing.
While writing this documentation, the author mentioned to Alonzo Johnson that there is no way right
now to quantify the amount of editing done on the data. Alonzo pointed out that the Liger processing
system does maintain an audit trail, but the indicator variables that show whether the data value has
been edited are not part of the export file that is used in the imputation system.

Analyze the 2014 JRFC paradata
The 2014 JRFC online data collection used the Census Bureau’s Centurion system. The author has the
paradata file from the Centurion system. The paradata file can be analyzed to figure out which
questions online respondents are avoiding, and which questions are problematic in terms of
respondents changing answers repeatedly. While 43.5% of the 2014 JRFC facilities used the online
reporting option, the method of return code is set once for a facility, and may not reflect the multiple
methods used to collect the data from the facility. In particular, we have evidence from another
criminal justice survey that some respondents have difficulty logging in and give up, or only provide a
handful of items before breaking off. Without analyzing the 2014 JRFC paradata, we have no idea how
much of a problem that is for juvenile residential facilities. See Dorinski and West (2014) for more
details.

2

For more details on American National Standards Institute/Federal Information Processing Standards (ANSI/FIPS)
codes, please see http://www.census.gov/geo/reference/ansi.html. Eliminating the Governments Division twodigit state code falls under Change Principle 1, using a standard metadata model as shared architecture. The 10
Guiding Change Principles are included in “The Future of Census Bureau Operations”, available at
https://www.census.gov/content/dam/Census/about/about-thebureau/PlansAndBudget/FutureofCensusBureauOperations_2013-04-25_v1.0_10ChangePrinciples.pdf.

Page 42 of 45
The 2014 JRFC paradata file may contain information about 1,056 juvenile residential facilities. Joseph
Moone’s ICES paper mentions research conducted in the late 1990s during the design phases for the
JRFC. In the first design phase, researchers conducted structured interviews with 40 facilities. In the
second design phase, researchers conducted cognitive interviewing with 18 facilities. Structured
interviews and cognitive interviews are worthy research techniques, but the paradata encompasses
many more facilities, and thus should be analyzed to see if respondents are having trouble with any of
the questions in the online data collection.

Include the mathematical statistician in project meetings
Historically, the mathematical statistician who handles the missing data has not been included in project
meetings. This means that the mathematical statistician has to update the imputation system
reactively, instead of proactively. Problems recognized during imputation processing can be difficult,
time-consuming, or expensive to fix. The mathematical statistician may be able to offer suggestions at
earlier stages in the project, when fixes would be easier, quicker, or cheaper.

Use data visualization more
The package of deliverables to OJJDP includes more than 150 tables. That is a lot of detail for someone
to review. Data visualizations might be easier to review quickly. Figure 7 shows the number of nontribal
juvenile residential facilities in the 50 states and the District of Columbia by the state juvenile civilian
population as of July 1, 2014. [The program plot_facilities_per_state_against_state_juv_pop.sas
produces the graphic.]

Consider additional edits for makeshift beds question
We recommend an edit to confirm reported occupied makeshift beds when the number of standard
beds is greater than the number of persons assigned to beds on reference day. We also recommend
edits that compare the data to prior year JRFC data.

Share data snapshot with facilities to help “sell” the data collections
As we noted in the 2013 CJRP imputation methodology documentation, respondents may not
understand the purpose of the data collection, nor how the data are used. There seems to be
understandable concern on the part of respondents to protect the data about the juvenile offenders
they house. However, respondents may not be aware that quite a bit of policy analysis is done at the
aggregate level. For example, see “Juvenile prison populations fall as states’ changes take effect”, by
Reid Wilson, published February 1, 2015 in the Washington Post and available online at
https://www.washingtonpost.com/blogs/govbeat/wp/2015/01/29/states-see-marked-drop-in-juvenileprison-populations-as-reforms-take-hold/.
We recommend doing more outreach with each data collection, letting the facilities know how the data
has been used and why it is important every time it is collected. OJJDP announced via email on October
20, 2015 the availability of a data snapshot based on the 2013 CJRP. Figure 8 shows the data snapshot.
The data snapshot would be an excellent addition to the survey package mailed to each facility. The
data snapshot should also be available on the online data collection tool.

Page 43 of 45

Figure 7. Number of nontribal facilities in the 50 states and District of Columbia by state juvenile population

Page 44 of 45

Figure 8. CJRP 2013 data snapshot

Page 45 of 45

References
Dilorio, Frank and Abolafia, Jeff. 2004. “Dictionary Tables and Views: Essential Tools for Serious
Applications”, Proceedings of the Twenty-Ninth Annual SAS® Users Group International Conference,
available online at http://www2.sas.com/proceedings/sugi29/237-29.pdf.
Dilorio, Frank and Abolafia, Jeff. 2006. “The Design and Use of Metadata: Part Fine Art, Part Black Art”,
Proceedings of the Thirty-first Annual SAS® Users Group International Conference, available online at
http://www2.sas.com/proceedings/sugi31/104-31.pdf.
Dorinski, S., West, H. 2014. What Paradata Can Tell Us About the Annual Survey of Jails. In JSM
Proceedings, Survey Research Methods Section. Alexandria, VA: American Statistical Association.
3154–3168. Available online at
http://www.amstat.org/sections/SRMS/Proceedings/y2014/Files/312871_90135.pdf.
Dorinski, Suzanne M. “Documentation of the Imputation Methodology for the 2013 Census of Juveniles
in Residential Placement”, memo dated June 26, 2015.
Dorinski, Suzanne. “Documentation of the Imputation Methodology for the 2012 Juvenile Residential
Facility Census”, memo dated March 13, 2014.
Other examples by Dorinski of SAS programming using metadata are online at
http://www.sascommunity.org/wiki/Presentations:Dorinski_Papers_and_Presentations.
Horwitz, Sari . “From Broken Homes to a Broken System”, published in the Washington Post on
November 28, 2014, and accessible online at
http://www.washingtonpost.com/sf/national/2014/11/28/from-broken-homes-to-a-broken-system/.
SC-EST2014-AGESEX-CIV: Annual Estimates of the Civilian Population by Single Year of Age and Sex for
the Unites States and States: April 1, 2010 to July 1, 2014. Washington, D.C.: U.S. Census Bureau,
Population Division. Release Date: June 2015.
Moone, Joseph. 2000. Innovative Information on Juvenile Residential Facilities. Washington, D.C.: U.S.
Department of Justice, Office of Juvenile Justice and Delinquency Prevention, NCJ# 188862. Available
online at https://www.ncjrs.gov/pdffiles1/ojjdp/fs200011.pdf.
Moone, J. 2000. Conceptualization and Measurement Issues for Surveys of Juvenile Facilities: Concepts,
Constructs, and the Impact of Empirical Research. In International Conference on Establishment Surveys
Proceedings. Alexandria, VA: American Statistical Association, 1226 – 1231. Available online at
https://www.amstat.org/meetings/ices/2000/proceedings/S16.pdf.
Wilson, Reid. “Juvenile prison populations fall as states’ changes take effect”, published February 1, 2015
in the Washington Post, available online at
http://www.washingtonpost.com/blogs/govbeat/wp/2015/01/29/states-see-marked-drop-in-juvenileprison-populations-as-reforms-take-hold/.