issued february 2004 special place/group quarters enumeration

39
Census 2000 Topic Report No. 5 TR-5 Issued February 2004 Special Place/Group Quarters Enumeration U.S. Department of Commerce Economics and Statistics Administration U.S. CENSUS BUREAU Census 2000 Testing, Experimentation, and Evaluation Program

Upload: others

Post on 07-Feb-2022

1 views

Category:

Documents


0 download

TRANSCRIPT

Census 2000 Topic Report No. 5TR-5

Issued February 2004

SpecialPlace/GroupQuartersEnumeration

U.S.Department of CommerceEconomics and Statistics Administration

U.S. CENSUS BUREAU

Census 2000 Testing, Experimentation, and Evaluation Program

The Census 2000 Evaluations Executive SteeringCommittee provided oversight for the Census 2000Testing, Experimentation, and Evaluations (TXE)Program. Members included Cynthia Z. F. Clark,Associate Director for Methodology and Standards;Preston J. Waite, Associate Director for DecennialCensus; Carol M. Van Horn, Chief of Staff; TeresaAngueira, Chief of the Decennial ManagementDivision; Robert E. Fay III, Senior MathematicalStatistician; Howard R. Hogan, (former) Chief of theDecennial Statistical Studies Division; Ruth AnnKillion, Chief of the Planning, Research and EvaluationDivision; Susan M. Miskura, (former) Chief of theDecennial Management Division; Rajendra P. Singh,Chief of the Decennial Statistical Studies Division;Elizabeth Ann Martin, Senior Survey Methodologist;Alan R. Tupek, Chief of the Demographic StatisticalMethods Division; Deborah E. Bolton, AssistantDivision Chief for Program Coordination of thePlanning, Research and Evaluation Division; Jon R.Clark, Assistant Division Chief for Census Design ofthe Decennial Statistical Studies Division; David L.Hubble, (former) Assistant Division Chief forEvaluations of the Planning, Research and EvaluationDivision; Fay F. Nash, (former) Assistant Division Chieffor Statistical Design/Special Census Programs of theDecennial Management Division; James B. Treat,Assistant Division Chief for Evaluations of the Planning,Research and Evaluation Division; and VioletaVazquez of the Decennial Management Division.

As an integral part of the Census 2000 TXE Program,the Evaluations Executive Steering Committee char-tered a team to develop and administer the Census2000 Quality Assurance Process for reports. Past andpresent members of this team include: Deborah E.Bolton, Assistant Division Chief for ProgramCoordination of the Planning, Research and EvaluationDivision; Jon R. Clark, Assistant Division Chief forCensus Design of the Decennial Statistical StudiesDivision; David L. Hubble, (former) Assistant DivisionChief for Evaluations and James B. Treat, AssistantDivision Chief for Evaluations of the Planning, Researchand Evaluation Division; Florence H. Abramson,Linda S. Brudvig, Jason D. Machowski, andRandall J. Neugebauer of the Planning, Research and Evaluation Division; Violeta Vazquez of theDecennial Management Division; and Frank A.Vitrano (formerly) of the Planning, Research andEvaluation Division.

The Census 2000 TXE Program was coordinated by thePlanning, Research and Evaluation Division: Ruth AnnKillion, Division Chief; Deborah E. Bolton, AssistantDivision Chief; and Randall J. Neugebauer andGeorge Francis Train III, Staff Group Leaders. KeithA. Bennett, Linda S. Brudvig, Kathleen HaysGuevara, Christine Louise Hough, Jason D.Machowski, Monica Parrott Jones, Joyce A. Price,

Tammie M. Shanks, Kevin A. Shaw, George A.Sledge, Mary Ann Sykes, and Cassandra H.Thomas provided coordination support. Florence H.Abramson provided editorial review.

This report was prepared by Florence H. Abramsonof the Planning, Research and Evaluation Division. Thefollowing authors and project managers preparedCensus 2000 experiments and evaluations that con-tributed to this report:

Decennial Statistical Studies Division:Nicholas S. AlbertiRoxanne FeldpaushKimball T. JonasTracey A. McNallyKevin J. Zajac

Planning, Research and Evaluation Division:Neala Stevens

The author would like to recognize the following quality assurance and operational assessment reportauthors: Carrie L. Johanson, Karen S. Medina,Broderick E. Oliver, and Sharon A. Schoch.

Greg Carroll and Everett L. Dove of the Admin-istrative and Customer Services Division, and WalterC. Odom, Chief, provided publications and printingmanagement, graphic design and composition, and edi-torial review for print and electronic media. Generaldirection and production management were providedby James R. Clark, Assistant Division Chief, andSusan L. Rappa, Chief, Publications Services Branch.

Acknowledgments

U.S. Department of CommerceDonald L. Evans,

Secretary

Samuel W. Bodman,Deputy Secretary

Economics and Statistics AdministrationKathleen B. Cooper,

Under Secretary for Economic Affairs

U.S. CENSUS BUREAUCharles Louis Kincannon,

Director

Census 2000 Topic Report No. 5Census 2000 Testing, Experimentation,

and Evaluation Program

SPECIAL PLACE/GROUPQUARTERS ENUMERATION

TR-5

Issued February 2004

Suggested Citation

Florence H. Abramson,Census 2000 Testing,

Experimentation, and EvaluationProgram Topic Report No. 5, TR-5,

Special Place/Group Quarters Enumeration,

U. S. Census Bureau,Washington, DC 20233

ECONOMICS

AND STATISTICS

ADMINISTRATION

Economics and StatisticsAdministration

Kathleen B. Cooper,Under Secretary for Economic Affairs

U.S. CENSUS BUREAU

Charles Louis Kincannon,Director

Hermann Habermann,Deputy Director and Chief Operating Officer

Cynthia Z. F. Clark,Associate Director for Methodology and Standards

Preston J. Waite, Associate Director for Decennial Census

Teresa Angueira, Chief, Decennial Management Division

Ruth Ann Killion, Chief, Planning, Research and Evaluation Division

For sale by the Superintendent of Documents, U.S. Government Printing Office

Internet: bookstore.gpo.gov Phone: toll-free 866-512-1800; DC area 202-512-1800

Fax: 202-512-2250 Mail: Stop SSOP, Washington, DC 20402-0001

U.S. Census Bureau Special Place/Group Quarters Enumeration iii

Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .v

1. Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11.1 Scope of the topic report . . . . . . . . . . . . . . . . . . . . . . . . . .11.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11.3 Historical background . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21.4 Special place/group quarters operational background . . . . .4

2. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .72.1 Evaluations E.5 and E.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . .72.2 ESCAP II Report 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .72.3 Decennial Management Division (DMD) operational

assessments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .72.4 Decennial Statistical Studies Division (DSSD)

quality assurance profiles . . . . . . . . . . . . . . . . . . . . . . . . . .72.5 Other source documents . . . . . . . . . . . . . . . . . . . . . . . . . . .7

3. Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .93.1 Topic report limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .93.2 Specific limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .93.3 General limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9

4. Major Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .114.1 Special place/group quarters inventory development . . . . .114.2 Enumeration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .124.3 Data capture and processing . . . . . . . . . . . . . . . . . . . . . . .18

5. Synthesis of Results and Recommendations for the 2010 Census . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .235.1 Revise group quarters definitions and classification

of group quarters type codes. . . . . . . . . . . . . . . . . . . . . . .235.2 Integrate housing unit and group quarters address list

development activities. . . . . . . . . . . . . . . . . . . . . . . . . . . .235.3 Begin planning the group quarters enumeration early. . . . .245.4 Continue research on service-based facility

enumeration operations. . . . . . . . . . . . . . . . . . . . . . . . . . .245.5 Track group quarters facilities and questionnaires

throughout the census. . . . . . . . . . . . . . . . . . . . . . . . . . . .245.6 Continue research on unduplicating people in

group quarters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .25

6. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .29

Contents

LIST OF TABLES

Table 1: Number of Special Places and Group Quarters Enumerated . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13

Table 2: Percent of Small and Large GQs Reinterviewed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17

Table 3: Non-ID Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19

Table 4: Persons in GQs from Household and Be Counted Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . .19

Table 5: Results of the SBE Unduplication of Data Captured Person Records . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .20

Table 6: E-Sample Duplicates to People Outside the A.C.E. Search Area in GQs . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21

iv Special Place/Group Quarters Enumeration U.S. Census Bureau

U.S. Census Bureau Special Place/Group Quarters Enumeration v

The Census 2000 Testing, Experimentation, and Evaluation Programprovides measures of effectiveness for the Census 2000 design,operations, systems, and processes and provides information on the value of new or different methodologies. By providing measuresof how well Census 2000 was conducted, this program fully sup-ports the Census Bureau’s strategy to integrate the 2010 planningprocess with ongoing Master Address File/TIGER enhancements andthe American Community Survey. The purpose of the report that follows is to integrate findings and provide context and backgroundfor interpretation of related Census 2000 evaluations, experiments,and other assessments to make recommendations for planning the 2010 Census. Census 2000 Testing, Experimentation, andEvaluation reports are available on the Census Bureau’s Internet siteat: www.census.gov/pred/www/.

Foreword

This page intentionally left blank.

1.1 Scope of the topicreport

The Special Place/Group Quarters(SP/GQ) Topic Report provides asynthesis of the Census 2000results, lessons learned, and rec-ommendations available on thedevelopment of the SP/GQ invento-ry, the enumeration of group quar-ters, and data processing opera-tions that were unique to thegroup quarters records.Specifically, the report answersquestions on how effective proce-dures were for developing theinventory of special places, howmany people were enumerated indifferent types of special places,how well procedures worked, andhow procedures might beimproved.

The report also provides a histori-cal perspective on SP/GQ enumera-tion from the 1990 Census and thetests leading up to Census 2000.Some operational background fromCensus 2000 is also provided.Most of the report is organizedaround three components of theSP/GQ operations: inventory devel-opment, enumeration, and datacapture and processing. The dis-cussion of data capture and pro-cessing is mostly limited to dupli-cation of people in group quarters.

There are additional topic reportsthat address subjects that havesome overlap with the informationcontained in the SP/GQ TopicReport:

• The Address List DevelopmentTopic Report provides informa-tion on operations used todevelop the Master Address File

(MAF). Operations used toinclude group quarters address-es on the MAF are discussed inthe SP/GQ topic report.

• The Coverage ImprovementTopic Report provides informa-tion on the coverage gains fromthe service-based enumerationoperations.

• The Coverage MeasurementTopic Report provides similarinformation to the SP/GQ topicreport on the evaluation ofgroup quarters persons duplicat-ed in housing units.

• The Data Collection Topic Reportprovides information on enu-meration of housing units.

• The Data Processing TopicReport provides information onGQ processing, including issueswith tracking GQ questionnairesand processing GQ responseswith Usual Home Elsewhere(UHE) addresses.

1.2 Introduction

The vast majority of United Statesresidents live in housing units(HUs). However, several millionpeople in the United States live ingroup situations, collectivelyknown as group quarters (GQs).While, overall, the decennial cen-sus was an HU based enumeration,unique operations were required tocompile the list of special placesand GQs and unique enumerationactivities were required to includeresidents of GQs in Census 2000.Within SP/GQ enumeration, therewere also unique operations toidentify locations and to include

persons without conventionalhousing for the service-based enu-meration (SBE).

Basic definitions of terms used inthe report:

Housing Unit (HU): An HU can be ahouse, an apartment, a mobilehome or trailer, a group of rooms,or a single room that is occupiedas a separate living quarters, or, ifvacant, is intended for occupancyas a separate living quarters.

Group Quarters (GQ): The CensusBureau defines GQs as placeswhere people live or stay otherthan the usual house, apartment,or mobile home. Examples of GQsinclude college and university dor-mitories, hospital/prison wards,and nursing homes. For purposesof evaluation, the GQ types wereclassified into eight broad cate-gories: Correctional Facilities,Juvenile Institutions, NursingHomes, Hospitals, Colleges andUniversities, Military Installations,SBE Facilities and Other GQs, andGroup Homes.

Special Place: A special place is anadministrative entity containingone or more group quarters wherepeople live or stay, such as a col-lege or university, nursing home,hospital, correctional facility, ormilitary installation or ship. A spe-cial place can include one or moreGQs. There can also be HUs at thespecial place, for example a col-lege president’s home on a collegecampus.

U.S. Census Bureau Special Place/Group Quarters Enumeration 1

1. Background

2 Special Place/Group Quarters Enumeration U.S. Census Bureau

1.3 Historical background

Since the Census 2000 operationalplan for GQ enumeration called formostly using the same basic 1990Census operations with some mod-ifications, most of the testinginvolving GQs during the decadeleading up to Census 2000 concen-trated on revising the approach toenumerating people without con-ventional housing. The followingdescribes the tests involving SBEenumeration. The major differ-ences between the 1990 Censusand Census 2000 SP/GQ inventorydevelopment and enumerationmethods are noted in Section 1.4.

In the 1990 Census, Shelter andStreet Night (S-Night) was an oper-ation designed to count personsliving in public and private shel-ters, pre-identified places of com-merce (e.g., bus and train sta-tions), and pre-identified streetlocations where people were visi-ble on the streets during the earlymorning hours. Advocates for thepeople experiencing homelessnessand some internal Census Bureaustudies pointed out that the proce-dures used in S-Night still resultedin an undercount of the populationthe method was supposed to cap-ture, since many of those peoplewould not have been “visible” tothe enumerators and would havehad no opportunity to be counted.This led to research into otherways to enumerate people experi-encing homelessness.

Beginning in late 1989, the CensusBureau began researching ways toimprove coverage of this popula-tion in the census by:

• conducting a small scale pilot-study of a day time service-based enumeration operation,

• evaluating the 1990 Census S-Night operations to identify

areas where procedures couldbe improved,

• funding a study to learn aboutlocal area methods,

• funding a contract to researchthe issue of sampling this popu-lation group, and

• conducting meetings withknowledgeable researchers, rep-resentatives from national andstate homeless coalitions,Federal agencies, and datausers.

Based upon the research, theCensus Bureau decided to test thefeasibility of a service-based enu-meration operation in the census.

In early 1994, Westat issued areport (Kalton et al, 1994) resultingfrom their work to assist theCensus Bureau in developing pro-cedures to be used in Census 2000for counting persons with no usualresidence. The report offered anumber of suggested methods,including using service providersas enumeration sources andrepeated enumerations at serviceproviders over a period of time.They suggested that the 1995Census Test be designed to collectinformation that could resolvemany of the unanswered questionsso that a decision could be reachedon the broad general approach tobe used in Census 2000. Specificrecommendations were:

• Conduct a complete enumera-tion of all shelters.

• Include soup kitchens, and pos-sibly other homeless outreachprograms, as the most promis-ing other sources for enumerat-ing persons without convention-al housing in the 1995 CensusTest.

• Examine the use of repeatedenumerations across time in the1995 Census Test and consider

embedding an experiment toexamine the effectiveness of dif-ferent time periods and differenttiming for the re-enumerations.

• Further examine the use of cap-ture-recapture methods for esti-mating the size of the popula-tion without conventionalhousing and assess the qualityof the data collected.

• Conduct a street enumerationon a sample basis in the 1995Census Test to determine thecoverage rate achieved for per-sons without conventional hous-ing.

The 1995 Census Test included thefirst attempt at a fundamentallydifferent approach to providingopportunities for enumerating per-sons without conventional hous-ing. These persons were enumer-ated at service locations such asshelters, soup kitchens, andhotels/motels. Using the SBEmethodology was expected toimprove coverage in the test areas.The 1995 Census Test methodolo-gy included one followup visit toshelters and soup kitchens to inter-view a sample of the clients anddetermine whether the data col-lected during the followup visitswould be sufficient to make reli-able statistical inferences aboutcoverage.

Results from the 1995 Census Testof the SBE methodology (Martin,1995) showed that:

• Enumerating clients at servicelocations should prove an effec-tive means of including an at-risk population in the censuscount.

• Operational procedures andquestionnaires need to be sim-plified.

• It may be beneficial to adapt theprocedures so that a common

U.S. Census Bureau Special Place/Group Quarters Enumeration 3

set of procedures could beimplemented in shelters andsoup kitchens.

• Multiple enumeration attemptsresulted in unacceptable levelsof respondent confusion.

Other Census Bureau research andtesting leading up to implementa-tion of the SBE in Census 2000included:

• 1993 - Griffin (1993) document-ed preliminary research on sam-pling and estimation of the pop-ulation experiencinghomelessness. The two parts ofthis research included 1) estima-tion in conjunction with sam-pling and unduplicating (i.e.,matching) persons enumeratedmore than once and 2) esti-mates of the total populationnot requiring matching.

• 1993 - The Census Bureaubegan a project to gather infor-mation on how communitieswere conducting their owncounts of people experiencinghomelessness. An aim of theproject was to assess whetherinnovative local methods couldbe adapted for later censusoperations.

• 1994 - Cognitive testing to eval-uate two forms specially pre-pared for the SBE initial and fol-lowup enumerations centeredon respondents’ comprehensionof the usual residence question,general readability of a self-administered form to the targetpopulation, the ability ofrespondents to recall where theyhad stayed over a period oftime, and the possible sensitivi-ty of the questions.

• 1996 - A small scale test in NewYork City was conducted to sim-plify procedures and the ques-tionnaire, to make them applica-ble to mobile food van sites as

well as soup kitchens, and tocorrect certain procedural prob-lems in the 1995 Census Test.

• 1998 - Census 2000 DressRehearsal - In addition to shel-ters, soup kitchens, and regular-ly scheduled mobile food vans,targeted non-sheltered outdoorlocations were enumerated aspart of the SBE for the first timein the dress rehearsal. BeCounted forms were distributedat targeted locations in thedress rehearsal. People enumer-ated on Be Counted forms whoindicated they had no addressand marked the “No Address onApril 18, 1998” box were includ-ed in the SBE universe. Also,respondents who gave a shelteras their usual residence on theBe Counted form became part ofthe SBE universe. Proceduresfor unduplicating people in theSBE universe were also devel-oped and tested in the dressrehearsal.

In addition to the above researchand testing, there was ongoingwork throughout the decade todevelop and improve operationalprocedures, questionnaires, andmultiplicity estimation for the SBE.Concurrent with the internalresearch, testing, and develop-ment, the Census Bureau spon-sored the following conferencesand meetings with outside stake-holders:

• 1993 - Research conference enti-tled “Towards Census 2000:Research Issues for ImprovingCoverage of the HomelessPopulation”

• 1994 - Expert Panel Meeting todiscuss the 1995 Census TestService-Based EnumerationOperation

• 1994 - Informational Meetingabout the “Service-Based

Enumeration Operation” plannedfor the 1995 Census Test

• 1998 - Expert Panel Meeting todiscuss the inclusion of“Targeted Non-shelteredOutdoor Locations” in theCensus 2000 Service-BasedEnumeration Program

Concurrent with the developmentof the SBE for Census 2000, theCensus Bureau worked with 12sponsoring federal agencies1 underthe auspices of the InteragencyCouncil on the Homeless to devel-op the 1996 National Survey ofHomeless Assistance Providers andClients. The Census Bureaudesigned and collected data for the1996 survey, which furnishedinformation about the providers ofhomeless assistance and the char-acteristics of the homeless popula-tion who use services, based on astatistical sample of 76 metropoli-tan and nonmetropolitan areas.

The 1996 survey was the firstnational study that produced infor-mation on the characteristics ofpeople participating in homelessassistance programs since a 1987study conducted by the UrbanInstitute. While using a similarmethodology to the 1987 study,the 1996 survey was based on alarger sample, collected more com-prehensive information, andincluded nonmetropolitan areas. Italso included a wider variety oflocations in order to reflect moreaccurately and fully the character-istics of homeless people who useservices nationwide. The 76 geo-graphic areas that were included in the national sample in 1996were comprised of the 28 largest

1 The 12 federal agency sponsors werethe Departments of Housing and UrbanDevelopment, Health and Human Services,Veterans Affairs, Agriculture, Commerce,Education, Energy, Justice, Labor, andTransportation, as well as the Social SecurityAdministration and the Federal EmergencyManagement Agency.

4 Special Place/Group Quarters Enumeration U.S. Census Bureau

metropolitan areas, 24 randomlyselected medium and small metro-politan areas, and another 24 ran-domly selected nonmetropolitanareas (small cities and rural areas).

1.4 Special place/groupquarters operationalbackground

Most of the information in this sec-tion was obtained from Jonas(2003). Information was alsoobtained from Gloster (2000),Halterman (2000), Schoch (2001),and Stackhouse (2000).

SP/GQ inventory development

Through 1998, the Census Bureaudid extensive development to iden-tify prospective special places,beginning with the cleanup of the1990 base file. The resultinginventory formed the basis of thefile known as the SP/GQ MasterFile. Additional special places fromother pre-Census operations wereadded to the SP/GQ Master File.

The SP/GQ Master File was themain repository of what wasknown about each special placeand each GQ. Additional files weredeveloped from it to supportupdating operations involving GQs.Updates were sent to theGeography Division (GEO) toupdate the MAF and assign MAFIdentification Numbers (MAFIDs).The SP/GQ Master File, updatedwith the addition of the MAFIDs,became the basis for later Censusoperations and enumeration.

For the 1990 Census, the localfield office staff updated, added,and deleted special places from theSpecial Place Master Listing basedon results from the Special PlacePrelist operation. The operationproduced too many changes to beprocessed in time for enumerationactivities. From that experienceand with improvements in technol-ogy, the Census Bureau planned

the Special Place FacilityQuestionnaire Operation to updateand correct the SP/GQ Master Filefor Census 2000. The Census2000 Special Place FacilityQuestionnaire operation was themeans for gathering informationabout GQs and HUs in each specialplace. For the different types ofspecial places, the interviewerasked the applicable questions tocollect information to identify whatGQs and HUs were at the specialplace, the hours of operation, themaximum number of clients or res-idents per GQ and the numberexpected in April 2000, theaddress of each GQ, and otherinformation required for planningthe enumeration.

For the 1990 Census, governmen-tal units were invited to participatein Precensus Local Review only inmailout/mailback areas. Theyreceived census maps and housingunit counts by block (but notcounts of special places or groupquarters). The Census 2000Special Place Local Update ofCensus Addresses (SP LUCA) opera-tion allowed local and tribal gov-ernments to examine the CensusBureau’s list of special places priorto GQ enumeration. They wereasked to identify special places noton the list, to provide their physi-cal location, mailing address, andother related information, and toassign special places to the Census2000 collection blocks in whichthey were located.

Similar to the 1990 Census LocalKnowledge Update operation, inthe Census 2000 Local KnowledgeUpdate operation, staff in eachLocal Census Office (LCO) reviewedthe lists to identify missing orduplicate special places.

During the Special Place AdvanceVisit operation (a new operationfor Census 2000), crew leaders vis-

ited special places on the addresslist to verify and correct informa-tion for the GQs and HUs at thespecial place, inform the contactperson about the enumerationoperations, assign missinggeocodes, and leave recruitingmaterials.

Group quarters enumeration

GQ enumeration was conductedApril 1, 2000 until May 6, 2000.GQ enumeration methods are dis-tinct from HU enumeration meth-ods. The household question-naires that work well for people inan HU are inappropriate for enu-merating larger groups of unrelat-ed people in places such as collegedormitories or nursing homes.The GQ questionnaires were per-son-based, rather than household-based, and they did not ask therelationship question or any hous-ing questions.

There were four main types of GQquestionnaires: the IndividualCensus Report (ICR), the IndividualCensus Questionnaire (ICQ), theMilitary Census Report (MCR), andthe Shipboard Census Report (SCR).The ICR was the form used to enu-merate the vast majority of the GQpopulation. The ICQ (a new formfor Census 2000) was used solelyfor enumeration at soup kitchensand regularly scheduled mobilefood vans. The MCR was usedsolely to enumerate armed forcespersonnel. The SCR was used toenumerate both military and civil-ian shipboard residents.

During the enumeration at eachGQ, enumerators obtained a list ofresidents, filled out a listing sheet,and distributed ICR packets to resi-dents. In some instances whereresidents could not fill out theforms themselves, enumeratorswere allowed to use administrativerecords to augment the process.They also used enumerator

questionnaires to enumerate resi-dents of HUs at the special place.

Certain GQs could request thatthey self-enumerate for the safetyof the enumerator or for the bene-fit or confidentiality of the resi-dents. The two most commontypes of self-enumerating GQswere hospitals and prisons.

Military enumeration operationscounted military personnelassigned to stateside militaryinstallations on Census Day.Dependents of the military person-nel living in family HUs on a mili-tary installation were enumeratedby the same methodology used forHUs in the area around the installa-tion.

The maritime/military vessels enu-meration operations counted indi-viduals assigned to ships based atAmerican home ports (military ves-sels) and American flag ships (mar-itime vessels). In 1990, vessels inthe 6th and 7th fleets were enu-merated with the overseas popula-tion. In 2000, service memberswere allowed to list a UHE address.If the respondent didn’t provide aUHE, they were counted at thehome port of the vessel. Servicemembers aboard vessels that werehome ported overseas were count-ed as part of the overseas popula-tion.

The transient night (T-night) enu-meration operation (conducted onMarch 31, 2000) enumerated peo-ple without a usual residence atlocations such as campgroundsand parks, commercial or publicfairs, carnivals, marinas, race-tracks, military hotels, and recre-ational vehicle (RV) parks. Theseare places where the residentstend to be highly transient.

SBE operations counted peoplewithout conventional housing who

could have been missed in the tra-ditional enumeration of HUs andGQs. People were enumerated atshelters (March 27, 2000), soupkitchens (March 28, 2000), regular-ly scheduled mobile food vans(March 28, 2000), and targetednon-sheltered outdoor locations(March 29, 2000).

Other differences between 1990Census and Census 2000 GQ enu-meration include modifications toT-Night enumeration and targetingcounties with high concentrationsof migrant and seasonal farmworkers at census time.

Data capture and processing

All GQ questionnaires were sent tothe data capture center at theNational Processing Center (NPC) inJeffersonville, Indiana. After all theGQ questionnaires were data cap-tured at the NPC and the captureddata transmitted to Census Bureauheadquarters, the SP/GQ MasterFile was updated.

The SBE enumeration involved anumber of opportunities for per-sons to be counted multiple times.An unduplication process removedduplicate enumerations from theCensus.

Respondents at certain types ofGQs could declare a UHE.2 That is,they could check a box to indicatethat the GQ was not their usualresidence and provide the addressof their usual residence. If theCensus was able to verify that an

HU was at the given address, thenthe respondent was counted atthat residence and not in the GQ.

Evaluations and assessments

There are three formal evaluationsthat were used in the preparationof this topic report, including onefrom the second Executive SteeringCommittee for Accuracy andCoverage Evaluation Policy (ESCAP II) series of reports.

The Census 2000 evaluationsinclude:

• Evaluation E.5, Group QuartersEnumeration

• Evaluation E.6, Service-BasedEnumeration

The ESCAP II report is:

• ESCAP II, Report Number 6,Census Person Duplication andthe Corresponding Accuracy andCoverage EvaluationEnumeration Status

In addition, planning documents,operational assessments, and qual-ity assurance profiles providedadditional information. Refer toSection 7, References, for a com-plete list of the documents thatwere used in the preparation ofthis report.

Sections 2 and 3 discuss the meth-ods used and the limitations,respectively. Section 4 summa-rizes the results, lessons learned,and recommendations contained inthe reports mentioned above,specifically intended to answerresearch questions on inventorydevelopment, the number of peo-ple enumerated in GQs, operationalissues, and recommendations foraddressing those issues. Section 5provides a synthesis of the majorrecurring themes contained in thereports mentioned above.

U.S. Census Bureau Special Place/Group Quarters Enumeration 5

2 The types of GQs in which residentswere not eligible to declare a UHE were pris-ons, juvenile institutions, hospital facilities,nursing homes, college dormitories, emer-gency shelters, and targeted non-shelteredoutdoor locations. All others who filled outGQ questionnaires were eligible to declare aUHE, including armed forces personnel,those enumerated at soup kitchens and reg-ularly scheduled mobile food vans, residentsof group homes, worker dormitories, civilianships, and religious GQs.

This page intentionally left blank.

2. Methods

The methodology of the programsinvolved in the source documents tothis topic report varied greatly bywhether the source was a statisticalevaluation, an operational report, ora quality assurance profile.

2.1 Evaluations E.5 and E.6

The evaluations of the GroupQuarters and Service-Based enumer-ations primarily provide descriptivestatistics of the operations, usingfiles that were already availablefrom Census 2000 or data files thatwere specifically created for theseevaluations. The files include:

• the SP/GQ Master File,

• the Hundred Percent CensusUnedited File (HCUF),

• the Hundred Percent CensusEdited File (HCEF),

• a data extract known as the Non-ID file which contained GQ per-son records that listed a UHE,and

• a data extract containing all SBEdata captured records, as well asBe Counted form records forwhich the person indicated they did not have an address on April 1, 2000.

The Group Quarters evaluation alsoused a statistical sample of GQsdesigned to estimate the level ofwithin GQ person duplication, notincluding military, correctional, andservice-based facilities.

2.2 ESCAP II Report 6

The ESCAP II evaluation of CensusPerson Duplication used data from

another ESCAP II report on personduplication (Mule, 2001) to evaluatehow the Accuracy and CoverageEvaluation (A.C.E.) coded E-Samplepeople (i.e., people enumerated incensus HUs in the A.C.E. sampleblock clusters) duplicated to peopleoutside the A.C.E. search area.

2.3 Decennial ManagementDivision (DMD) operationalassessments

The DMD operational assessmentswere the main vehicle for document-ing Census 2000 information into aset of comprehensive, integratedreports. Participating divisions wereasked to take an active role bypreparing or supplying initial assess-ments of relevant aspects of particu-lar operations or functional pro-grams. The initial assessmentsmade use of available sources(debriefing results, observationreports, formal and informationreports from staff and contractors,memoranda, and the like) andresources within each participatingdivision to assess the effectivenessof each operation or function as wellas to identify lessons learned andrecommendations for the next cen-sus. A team of representatives fromDMD and the participating divisionssubsequently organized and ana-lyzed these initial assessments,along with other sources of informa-tion, to produce comprehensiveassessments for the entire program.The development of the recommen-dations focused on the individualoperations, without an attempt toassess the implications across theentire census process.

The DMD operational assessmentsused in this topic report include:

• SP/GQ Inventory Development

• SP/GQ Enumeration Operations

• 1998, 1999, and Special PlaceLocal Update of CensusAddresses and New ConstructionPrograms

2.4 Decennial StatisticalStudies Division (DSSD)quality assurance profiles

The DSSD Quality Assurance (QA)Profiles provide assessments ofCensus 2000 operations based ondata collected from the QA pro-grams instituted for those opera-tions. Many of the QA programsinvolved relisting or reinterviewingprocedures, usually conducted ona sample basis. The DSSD QA pro-files used in this topic reportinclude:

• Profile of the Military GroupQuarters Address Listing QualityAssurance Operation

• Profile of the Census 2000Group Quarters ReinterviewOperation

2.5 Other sourcedocuments

Some other source documents wereused as input to this topic report.They include the following opera-tional planning documents: SP/GQInventory Development, SpecialPlace LUCA, GQ Enumeration,Service-Based Enumeration, MilitaryInstallation Enumeration, andQuestionnaire Assistance Center andBe Counted Program Master Plans.

U.S. Census Bureau Special Place/Group Quarters Enumeration 7

This page intentionally left blank.

3. Limits

3.1 Topic report limits

The research questions mentionedin Section 1.1 are answered to theextent that information was avail-able in the evaluation and assess-ment reports. Some questionscould not be fully answeredbecause of cancelled evaluationsand/or reports not available at thetime this report was written.

Most of the information on filedevelopment and enumerationactivities is derived from opera-tional assessments, rather thanfrom empirical research. Theassessments are useful for futureplanning in that they documentlessons learned from Census 2000and provide recommendations foroperational planning and testingfor the 2010 Census. However,although the assessments wereintended to be comprehensive,many of the statements made inthe reports do not include informa-tion that would help in understand-ing exactly what was learned orwhy a recommendation was pre-sented.

3.2 Specific limits

The evaluation reports list the lim-its of the individual studies. Theyare briefly summarized here.

3.2.1 GQ enumeration

Individual GQ questionnaires con-tained a space where enumeratorsrecorded the GQ identificationnumber (ID) and the person num-ber at the time of enumeration.This tracking information was notused to track the progress of indi-vidual forms from enumeration todata capture.

The limitations of the systemdesigned to track special placesand GQs from the beginning of theSpecial Place Facility Questionnaireprocess through tabulationreduced the ability to compare thesame special place before and afterthe Special Place FacilityQuestionnaire Operation.

Furthermore, if special places andGQs were deleted and later re-added during enumeration, theyreceived new ID numbers whenthey were re-added. As a result,comparing snapshots of the SP/GQMaster File at different times didnot allow a record of a specialplace at different points in time tobe identified as the same place, ifthe special place was deleted andre-added during enumeration.

3.2.2 SBE enumeration

A small percentage of the total GQquestionnaires were not data cap-

tured because the questionnairescould not be identified with a spe-cific GQ as a result of blank,incomplete or erroneous GQ IDs.Some portion of these affected theSBE universe. Questionnaires notcaptured would affect the popula-tion counts for SBE locations butthere is no information about thenumber or source of these missingquestionnaires.

We accepted a UHE address frompeople enumerated at soupkitchens and regularly scheduledmobile food vans. The data avail-able to the evaluation do notinclude information about UHEaddresses reported by SBE respon-dents.

3.2.3 Census person duplication

The ESCAP II, Report Number 6does not separately examine theissue of movers. It doesn’t meas-ure the amount of duplication dueto people moving during the peri-od of the enumeration.

3.3 General limits

In addition to the specific limita-tions, there were opportunities forfield and processing operationaldeviations that could affect thedata in the evaluation reports, theoperational assessments, and thetopic report.

U.S. Census Bureau Special Place/Group Quarters Enumeration 9

This page intentionally left blank.

4.1 Special place/groupquarters inventorydevelopment

Most of the information in this sec-tion came from Schoch (2003a)and Johanson and Oliver (2003).

Schoch (2003a) reported threemajor areas of success in the over-all Census 2000 SP/GQ inventorydevelopment process:

• Census 2000 was the first cen-sus in which the SP/GQ MasterFile represented the creationand maintenance of a linkbetween the special place andits associated GQs and theestablishment of a centralizedrepository for the completedecennial inventory of GQs.

• The Special Place FacilityQuestionnaire provided a struc-tured and consistent approachto obtain information about theGQs.

• The Special Place Advance Visitafforded a “last minute” oppor-tunity to visit the facility andverify, update, and correct infor-mation prior to enumeration.

Aside from these major areas ofsuccess, there were some aspectsof the development process wherethere were noteworthy lessonslearned and recommendations.These areas were:

• Definitions and Classifications

• Duplication

• Geocoding

Issues and recommendations con-cerning definitions are discussed in

Sections 4.1, 4.2 and 5.1.Duplication issues and recommen-dations are discussed in Sections4.3 and 5.6. Geocoding issues arediscussed below.

4.1.1 Initial file development

The creation of the Census 2000SP/GQ Master File began with thecleanup of the 1990 file. Thisincluded the creation of the linkagebetween the special place and itsassociated GQs. In order to updatethe 1990 file, most of the samesources used for the 1990 opera-tion were contacted for each typeof special place. Multiple sourceswere contacted for some specialplace types that contributed toduplication of facilities within thefile. Due to a poor response fromsome of the initial sources, therewere deficiencies in certain cate-gories (particularly the migrantworker camp and group homeinventories).

4.1.2 Special place facility questionnaire

The Special Place FacilityQuestionnaire was designed toprovide a systematic method toverify and update informationabout known special place facilitiesand all associated GQs and HUs.The initial Special Place FacilityQuestionnaire information was col-lected through a ComputerAssisted Telephone Interview(CATI) and the followup informa-tion was collected through fieldpersonal visit.

Although the Special Place FacilityQuestionnaire allowed for a morestructured and consistent approach

for obtaining, correcting, and veri-fying information about specialplaces and their associated GQs,the overall design of the SpecialPlace Facility Questionnaire wascomplex, confusing, and not user-friendly. The amount of informa-tion initially collected was toodetailed and lengthy for a tele-phone interview. This was espe-cially true for large facilities con-taining multiple GQs. Also, tryingto obtain geocoding informationby telephone was very difficult.

A major component of the trainingand procedures was the geocodingof the physical location of the spe-cial place and its associated GQsand HUs. Although a detailed sec-tion on geocoding procedures waswritten and presented in the train-ing sessions, the sizable number ofproblems suggests this processwas not effective. The failure toconsistently and correctly assigngeocodes created problems fromthe time of enumeration andresulted in the geographic misallo-cation of GQs in data tabulations.

4.1.3 Special place LUCA, localknowledge update, and specialplace advance visit

There were three pre-enumerationoperations to enhance and refinethe SP/GQ inventory prior to enu-meration.

• Special Place LUCA was anopportunity for participatinggovernmental units to reviewthe existing special place inven-tory list and provide feedback tothe LCOs for additions, dele-tions, and corrections.

U.S. Census Bureau Special Place/Group Quarters Enumeration 11

4. Major Findings

12 Special Place/Group Quarters Enumeration U.S. Census Bureau

• Local Knowledge Update was anoperation where the specialplace staff at the LCO reviewedthe SP/GQ inventory list forcompleteness and accuracyusing their local knowledge ofthe special place facilities andlocations in their area.

• Special Place Advance Visitrequired the special place LCOstaff to visit each special placeon the final inventory list to ver-ify, update, and correct all infor-mation about the special placeand the associated GQs andHUs.

The Special Place Advance Visitwas successful both in terms oftiming (right before GQEnumeration) and information veri-fication.

Although the overall LUCA pro-gram was successful in fosteringpositive partnerships, exceedinglocal and tribal government partici-pation, and contributing newaddresses, deleting addresses, andmaking address corrections, one ofthe lessons learned was that spe-cial place addresses should havebeen included with housing unitaddresses in LUCA.

Although these three operationswere designed as distinct opera-tions, time constraints of the pre-enumeration schedule forced themto overlap, reducing the impactand overall effectiveness of eachindividually.

4.1.4 Military group quartersaddress list development

As reported in Johanson and Oliver(2003), the Census Bureau con-ducted an operation called MilitaryGQ Address Listing in 1999 to veri-fy and update the database of GQson military installations. For enu-meration purposes, Military GQsincluded barracks, unaccompaniedofficer quarters, disciplinary bar-

racks, and military hospitals.(Military hotels and campgroundswere included in T-Night opera-tions.) To ensure the quality of theaddress information, a QA programwas instituted. Due to the relative-ly small workload and because themilitary installations are geographi-cally scattered, address listers per-formed QA on their own work.They selected a sample of the GQson the installation and for eachselected GQ, they verified theaddress information and geocod-ing. They also performed a cover-age check by inquiring about thefunction of the buildings to theimmediate left and right of eachsampled GQ.

There were 669 military installa-tions for which QA data werereceived. Of these, 659 (98.5 per-cent) had no critical errors (i.e., notranscription, map spot, orgeocode errors). In the ten mili-tary installations with errors, therewere 57 errors: 15 transcriptionerrors, 24 geocoding errors, and18 map spot errors. The 57 errorscame from four out of the 222 lis-ters (1.8 percent of the listers).

4.1.5 Recommendations

The following recommendationswere made by Johanson and Oliver(2003), Medina (2003), and Schoch(2003a and 2003b):

Reevaluate the role of the SpecialPlace Facility Questionnaire.Develop and design a streamlineduser-friendly questionnaire, whichincorporates terminology and clas-sifications consistent with the cur-rent industry. Limit the amount ofinformation collected through tele-phone contact.

Make use of Global PositioningSystem or other current technologyto satisfy geocoding requirements.

Extend the Special Place AdvanceVisit operation to include a tele-

phone followup to the facility clos-er to the time of enumeration. Thefollowup should include schedulingthe enumeration appointment andfinal verification of the contact per-sons and expected populations ofall the associated GQs.

Reevaluate the Local KnowledgeUpdate and SP LUCA operations bylooking at alternative points in theprocess to implement. Integratethe SP LUCA with the other LUCAoperations.

Reevaluate and design a method tointegrate the SP/GQ inventorydevelopment into the overall MAFdevelopment process.

Use more controls to track theSP/GQ inventory development QAoperation and obtain more infor-mation to be able to determine thecause of errors.

Maintain the military installationlists between censuses by monitor-ing the status of closings, down-sizing, etc.

4.2 Enumeration

Schoch (2003b) reported threeareas of success in the overallCensus 2000 SP/GQ enumeration:

• Service-Based Enumeration wasan improvement over previousdecennial operations to includepeople without conventionalhousing.

• Group Quarters Reinterview pro-vided quality control and was apositive addition to the SP/GQenumeration operations.

• Military Enumeration planningwas supported by a cooperativeeffort between the military andthe Census Bureau and provideda foundation for a smooth, time-ly operation. The Military VesselEnumeration also highlightedthe importance of accessible

liaisons from the Navy andCoast Guard.

Schoch (2003b) also reported somechallenges:

• T-Night Enumeration raisedquestions if this operation reallybelongs under SP/GQ operationsor should be covered by HUenumeration. Although the enu-meration was part of the SP/GQenumeration operations, thesites/slips/units with personswith no usual home elsewherewere enumerated as HUs ratherthan GQs.

• GQ Enumeration identified newissues for future consideration,including privacy/confidentialityissues3, gaining cooperationfrom managers at the appropri-ate levels of large specialplaces, and properly linking

individuals to their associatedGQs.

4.2.1 The GQ universe

Jonas (2003) reported the overallcomposition of the GQ universe asenumerated in Census 2000.

Approximately 7.8 million peoplewere tabulated in places coveredby the GQ universe in Census2000. Most of these people livedin colleges, prisons, or nursinghomes.

Roughly 78 percent of the specialplaces enumerated consisted ofonly one GQ. Over 98 percentcontained fewer than ten GQs.The largest proportion of specialplaces with more than 50 GQsincluded military installations, col-leges and universities, and correc-tional facilities. The smallest pro-portion of special places with morethan 50 GQs included nursinghomes and SBE facilities and other GQs.

Forty percent of special places hadfewer than ten residents and 61percent had fewer than 25 resi-dents. These were mostly grouphomes, SBE facilities, and othersmall GQs. The 40 percent of spe-cial places with fewer than ten res-idents accounted for 2.3 percent ofthe GQ population.

Even though enumerators wereallowed to use administrativerecords to augment the processonly in cases where residentscould not fill out the forms them-selves, Jonas (2003) found thatalmost half (48.9 percent) of all GQquestionnaires were filled out fromadministrative data and that mostGQ questionnaires at correctionalfacilities, hospitals, nursing homes,and group homes were filled outfrom administrative data. OtherGQ questionnaires were eitherfilled out by a respondent (24.8percent) or by an enumerator (9.7percent). The item indicatingmethod of response was blank orinvalid for 16.6 percent of the GQquestionnaires.

4.2.2 The SBE universe

McNally (2002) reported the fol-lowing:

• There were 14,817 SBE sites inCensus 2000. More than half(51 percent) of the locationswere shelters.

• There were a total of 258,728person records data capturedfrom shelters, soup kitchens,regularly scheduled mobile food vans, and targeted

U.S. Census Bureau Special Place/Group Quarters Enumeration 13

3 In Census 2000, some administratorsof group quarters cited both Federal andstate laws that prevented them from allow-ing census takers access to records or entryonto the premises to interview inhabitants.Issues were raised with regard to education-al and medical records, nursing homes, anddrug and alcohol abuse facilities. Censusenumerators eventually got into every facili-ty, but there is a need to know what Federaland/or state laws exist so that the CensusBureau can proactively develop measures toovercome this problem for the 2010 Census.

After Census 2000, the Census Bureaufunded a contract with Privacy Journal to 1)research and identify Federal and state lawsthat could be interpreted to prevent accessby census enumerators to the physical facili-ties of group quarters or to the records oninhabitants of the group quarters, 2)describe each such law, 3) consult withCensus Bureau specialists to resolve anyquestions, 4) organize descriptions of lawsby type, and 5) prepare interim and finalreports.

The resulting report distinguishesbetween access to facilities and access torecords. The laws and requirements onaccess to records are more highly developedthan laws or requirements on access to facil-ities. In short, Privacy Journal's researchuncovered only a few instances where thereis a law that could be cited to deny physicalaccess to census workers. Proprietors ofgroup quarters may have been reacting to ageneralized concern about privacy or to ageneral memory that there are laws on thebooks preventing access.

When confronted with denials of access,Census workers should be able to consultwith a legal advisor who has on file a reportlike the one prepared by Privacy Journal.

Table 1.Number of Special Places and Group Quarters Enumerated4

SP/GQ typeSpecial places Group quarters Population

Number Percent Number Percent Number Percent

1. Correctional facilities . . . . . 5,420 5.4 15,775 8.2 1,993,302 25.52. Juvenile institutions . . . . . . 2,440 2.4 6,335 3.3 129,132 1.73. Nursing homes . . . . . . . . . . 21,051 21.0 29,736 15.5 1,727,811 22.14. Hospitals. . . . . . . . . . . . . . . . 3,552 3.5 9,289 4.8 237,597 3.05. Colleges and

universities . . . . . . . . . . . . . 3,528 3.5 23,842 12.4 2,066,302 26.46. Military installations . . . . . . 916 0.9 6,104 3.2 356,354 4.67. SBE facilities and other

GQs . . . . . . . . . . . . . . . . . . . 38,783 38.7 56,092 29.1 854,435 10.98. Group homes. . . . . . . . . . . . 24,668 24.6 45,113 23.5 460,474 5.9

Total . . . . . . . . . . . . . . . . . . 100,358 100.0 192,286 100.0 7,825,407 100.0

Source: Jonas (2003), Tables 4.1a and 4.1b.

4 The data in Table 1 include countsfrom Puerto Rico.

non-sheltered outdoor locations.Most of the data captured per-son records (90 percent) werefrom shelters, soup kitchensand regularly scheduled mobilefood vans.

• Almost all (99 percent) of thedata captured person recordshad at least two or more datacharacteristics (name, sex, ageand/or date of birth, Hispanicorigin, or race) and were classi-fied as data defined.

• Nearly three-fourths (72 percent)of the SBE person records hadall five data characteristics com-pleted. This compares to 88percent for the HU personrecords with all five data charac-teristics reported.

Multiplicity estimation for SBE wasintended to use service usageresponses to adjust enumerationcounts for persons who use servic-es but not on the day of enumera-tion. During the SBE data collec-tion operation, respondents wereasked the number of times in aweek that they use that type offacility.

• In shelters, respondents wereasked “Including tonight, howmany nights during the pastseven nights did you stay in ashelter?”

• In soup kitchens and regularlyscheduled mobile food vans,respondents were asked“Including today how many daysduring the past seven did youreceive a meal from a soupkitchen or mobile food van?”

The responses to these questionscould be from 1 to 7 or a nonre-sponse. A person who only used aservice one night out of the sevennights in a week would be given aweight of seven since they onlyhad one chance in seven to beenumerated on the day selected

for enumeration. Conversely, a

person who used a service all

seven nights in a week would be

given a weight of one since they

would be enumerated no matter

what day was selected for enumer-

ation.

Corrected census data would have

had different counts of total popu-

lation using SBE facilities than the

uncorrected data if we had joined

results from the SBE multiplicity

estimation with results from the

actual Census 2000 SBE enumera-

tion. Specifically, the corrected

data would have used SBE multi-

plicity estimation to estimate the

number of people who use servic-

es but not on the day of enumera-

tion.

A decision was made NOT to use

SBE multiplicity estimation to esti-

mate the number of people who

use services but not on the day of

enumeration. The usage respons-

es had both a high nonresponse

rate and, particularly in shelters, a

very high level of response bias.

Accurate responses to the usage

question are critical for multiplicity

estimation. For additional informa-

tion on the multiplicity estimator

and the decision to not use it, see

Griffin, 2001.

4.2.3 Field operations

The issues and lessons learned

outlined in sections 4.2.3.1,

4.2.3.2, and 4.2.3.3 come from

Schoch (2003) and the QA results

in section 4.2.3.4 come from

Oliver (2002).

4.2.3.1 Planning

Schoch (2003) reported that even

though the enumeration plans

were tested in the field, there were

changes made after the tests and

before Census 2000.

• The late identification and solid-ification of requirements impact-ed the understanding of thedetails by headquarters staff forvarious operations. This wasreflected in the inconsistency ofthe information on SP/GQs indifferent field operational mate-rials.

• Delays in the final operationaland procedural requirementsdelayed the completion of allrequired enumeration materials(training guides, job aids, etc.).This caused a delay in kit prepa-ration and delivery, which sub-sequently diminished the timeallowed for preparation ofSP/GQ enumeration operationsin the LCOs.

4.2.3.2 Staffing, training, andoffice operations

In the LCO, the operations weremanaged by the Special PlaceOperations Supervisor (SPOS). TheSPOS trained the crew leaders andthen the crew leaders trained theenumerators. Each operation hadits own training guide, workbook,checklist, and job aid color codedby operation to help identify eachoperation’s materials. Among thelessons learned are:

• The SPOS was hired too late(late 1999) to adequately pre-pare for the job. A SPOS withlittle or no experience trainedand managed people on detailedoperations with little or no over-sight or assistance from a super-visor.

• Problems also occurred afterSP/GQ enumeration when ques-tions involving SP/GQs contin-ued through the different fieldfollowup operations and theSPOS position had already beenterminated. The LCOs had difficulty handling theserequests unless the SPOS was

14 Special Place/Group Quarters Enumeration U.S. Census Bureau

still working in the office insome other position and wasavailable to help.

• The number of work hours andweeks in the standard recruitingscripts did not apply to theSP/GQ operations.

• The different enumeratorrequirements for each SBE oper-ation were difficult to maintainand track. For example, someoperations required two-mem-ber teams and others requiredseven-member teams.

• SBE operations had the most dif-ficulty hiring and training ade-quate, knowledgeable staff forconsecutive one day operations.

• Special place staff left beforecompleting their special placework due to overlapping fieldoperations. For example, train-ing for nonresponse followupbegan prior to the completion ofGQ enumeration. Many of theexperienced GQ enumeratorswere taken from the specialplace operation to work on non-response followup so theywould have an opportunity towork longer. Additionally, theAccuracy and CoverageEvaluation (A.C.E.) operationpaid their enumerators $1.00more per hour. Some GQ enu-merators transferred to theA.C.E. operation for higher payand more hours before complet-ing their special place work.

• The training did not prepareenumerators to deal with thepublic. Enumerators had to con-tact management personnel atdifferent levels at special placesas well as interview individualsin unique situations encoun-tered during the SP/GQ enumer-ation operations.

4.2.3.3 Enumeration operations

The SP/GQ enumeration operationswere tightly scheduled, often over-lapping the training of one opera-tion with the field work of another.Each operation had its own set ofinstructions tailored to the popula-tion it was designed to enumerate.The documented lessons learnedinclude:

• Types of living quarters such asassisted living facilities, person-al care and retirement homes,and boarding schools wereoften misclassified as GQs dur-ing Census 2000. The GQ defi-nitions did not specifically listthese as exclusions. This result-ed in confusion in the field.

• It was difficult to identify certainGQs because they often “looklike” regular HUs (e.g., smallgroup homes and off-campus,college owned or leased apart-ment buildings). This con-tributed to the creation of dupli-cates between the HU and GQuniverses.

• The self enumeration optioncaused problems in the field.This option was presented tofacilities during the Special PlaceAdvance Visit, but was not com-pletely understood by the facili-ty contact person at that time.Many facilities chose this optionbecause they planned to useadministrative records. Oncethey received the actual train-ing, they realized they wereexpected to conduct an enumer-ation operation.

• SBE had four separate compo-nents designed to count peoplewithout conventional housingwho may be missed in the tradi-tional enumeration of HUs andGQs. The four operations wereShelter Enumeration, SoupKitchen Enumeration, Regularly

Scheduled Mobile Food VansEnumeration, and Targeted Non-sheltered Outdoor LocationsEnumeration. These were chal-lenging operations to includepersons without conventionalhousing in the Census. The tim-ing for all four SBE Operationsand the required training ses-sions over three consecutivedays in addition to T- NightOperations at the end of thesame week exhausted staff atthe LCOs, Regional CensusCenters (RCCS), and headquar-ters.

• The use of the ICR (designed fora respondent to complete) dur-ing the Targeted Non-shelteredOutdoor Locations Enumerationdid not adapt well to use as anenumerator interview form.

• There were complaints fromsome soup kitchens that werenot open on the one day desig-nated for that particular enu-meration.

• Transient Night (T-Night)Enumeration was part of theSP/GQ Enumeration operations.However, the sites/slips/unitswith persons with no UHE wereenumerated as HUs rather thanGQs.

• There was not enough time tothoroughly prepare GQ enumer-ators on how to administer thehousehold questionnaire for aone night operation.

• Conducting T-Night Enumerationon a Friday evening increasedthe amount of screening enu-merators had to do becausemany individuals were awayfrom home for the weekend atT-Night locations. For individu-als who had a usual home else-where but who were at the T-night location at the time ofenumeration, the enumerator

U.S. Census Bureau Special Place/Group Quarters Enumeration 15

had to give them a noticeexplaining they would be enu-merated at their primary resi-dence and then appropriatelycheck the UHE column on thelisting sheet. After enumera-tion, the enumerator had to tallythe number of checks in theUHE column and enter the totalon the GQ Enumeration Record.

• Military Group QuartersEnumeration was a combinedplanning effort by the CensusBureau and liaisons from eachbranch of the Armed Forces.

• FLD and GEO required a com-plete installation list but for dif-ferent purposes and require-ments which caused problemsin initial installation list develop-ment.

• The concept of UHE was notclearly understood by militarypersonnel.

In July 2000, after GQ data capturewas completed, two special fol-lowup operations were undertakento review and repair problems withthe GQ enumeration counts. Onetelephone operation was conduct-ed out of the RCCs to contact‘refusals’ that would not allow enu-meration to take place at theirfacilities. The other telephoneoperation was conducted out ofthe NPC to contact facilities whosedata capture population of a GQwas substantially lower thanexpected. Both of these operationsasked the contact persons at facili-ties what their Census Day popula-tion had actually been. If the facili-ty provided a count of itspopulation on April 1, 2000, thatcount was accepted as definitivefor that facility, unless the countwas lower than the actual numberof forms captured for that facility.Together, these operationsaccounted for imputing over100,000 persons in the GQ uni-

verse that were not included in thedata capture count.

Another enumeration problemidentified during processing wasthat 141,055 questionnaires weremissing a GQ ID number or had aninsufficient GQ ID number so thatthey could not be matched to alegitimate GQ ID on the SP/GQMaster File. These accounted forabout 1.7 percent of the total GQquestionnaires. About 61 percentwere found to be shifts or transpo-sitions of legitimate GQ ID num-bers and were ultimately matched.However, the remaining 39 percent(0.7 percent of the total GQ ques-tionnaires) could not be matched.

4.2.3.4 QA reinterview

As reported in Oliver (2002), theCensus Bureau conducted a com-prehensive QA program that tar-geted all GQs except those involv-ing the military, SBE facilities, andcivilian ships. About 89 percent ofall GQs were included in this QAprogram. Different QA procedureswere used for the other 11 percentof the GQs.

The QA program involved threeparts:

• A clerical completeness checkcompared the number of ICRsreceived from a given GQ to thenumber of residents listed.

• An automated population esti-mate check flagged those GQswhere the number of checked-inICRs did not fall within ten per-

cent of the estimated CensusDay population (obtained fromthe Advance Visit).

• The reinterview program target-ed all large GQs (populationgreater than or equal to 100)and any small GQs (populationless than 100) that failed thepopulation estimate check.Table 2 provides the percent ofGQs that were actually includedin the reinterview program. Asstated in Section 4.2.1, 40 per-cent of special places had fewerthan ten residents. Assumingan equal or greater percentageof GQs had an expected popula-tion of ten or less, than these“small” GQs would fail the popu-lation estimate check if the actu-al count was off by just one.This could account for the largepercent of small GQs included inthe reinterview sample. Oliver(2002) indicates possible rea-sons why less than 100 percentof the large GQs were included.

The overwhelming majority of theGQs reinterviewed passed the rein-terview process (that is, the GQcontact person verified that anenumerator visited the GQ andobtained the correct Census Daycount). Overall, 96.3 percentstateside and 97.7 percent inPuerto Rico passed the reinterviewprocess.

The data also suggest that certaintypes of GQs (e.g., Colleges andUniversities and Nursing Homes)had higher rework percentagesthan other GQ types.

16 Special Place/Group Quarters Enumeration U.S. Census Bureau

Table 2.Percent of Small and Large GQs Reinterviewed(In Percent)

Size of GQ Stateside Puerto Rico

Small . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30.3 42.8Large. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62.9 59.3

Source: Oliver (2002) Tables 4.2b and 5.2b.

4.2.4 Lack of tracking informationfor GQ questionnaires

GQ s were tracked through theenumeration and data captureprocess and counts of the GQquestionnaires were recorded oneach GQ control sheet at severalpoints between enumeration anddata capture, but the individualquestionnaires were not tracked.

One of the problems resulting fromthis was that the official populationcounts for some GQs were indirect-ly derived. In August 2000, aneffort was made to reconcile thecount of questionnaires checkedout of the LCO and the number ofquestionnaires data captured forthat GQ. To avoid problemscaused by forms from one GQ at aspecial place being incorrectlycounted with those from anotherGQ at the same special place, thereconciliation was done at the spe-cial place level. The reconciliationconsisted of taking the larger ofthe total LCO checkout populationand the total data capture popula-tion for each special place as the“official population” for that specialplace. Reconciliation of the differ-ing counts resulted in the imputa-tion of over 100,000 personrecords.

The inability to track individualquestionnaires also resulted in nothaving an exact count of personsadded to or subtracted from theSP/GQ Master File. The GQ popula-tion count decreased between thetime of the enumeration and thefinal Census count. Most of thisdecrease was a result of militarypopulation removed from the GQuniverse if they listed a UHE ontheir questionnaire. (See Section4.3.1 for more detail on MilitaryUHEs.) However, since individualGQ questionnaires were nottracked, the only information avail-able concerning the decrease is the

net gain or loss for each GQ. (Forexample, if a GQ gained two peo-ple and lost seven, all that isknown is that there was a net lossof five people.)

4.2.5 Recommendations

The following recommendationswere made by Jonas (2003),McNally (2002), Oliver (2002), andSchoch (2003b):

Begin planning SP/GQ Enumerationoperations early in the decade soall plans are completed and inplace to allow for consistency andtimely completion of all the associ-ated materials (job aids, work-books, training guides, etc.).

Research and revise GQ definitionsto include how to classify assistedliving facilities, personal carehomes, and retirement homes.

Meet with advocates and expertsto update the GQ definitions byusing current terminology.Although the definitions and classi-fications were researched and test-ed, there were some inherentweaknesses. The definitions andterminology used by the CensusBureau were not always consistentor current with that of other agen-cies, local governments, and thegeneral public.

Reduce duplication between theHU and GQ address lists. Manysmall GQs are indistinguishable inappearance from single-family resi-dences. As a result, some foundtheir way into both the HU and GQuniverses before being identifiedas the same place.

Establish a special place team atheadquarters to keep SP/GQEnumeration operational planningand implementation moving for-ward throughout the decade. Thiswill provide a much needed consis-tent foundation to develop, ana-

lyze, and maintain these variousoperations.

Evaluate whether there are benefitsto be gained in using different pro-cedures for different categories ofGQs in 2010, in building the SP/GQinventory, in the enumerationitself, and in post-enumeration pro-cessing.

Investigate alternative methods ofenumeration such as onlineresponse, proxy interviews, elec-tronic files, and use of administra-tive records other than as a lastresort for certain types of GQs.Research the quality of administra-tive records for GQs.

Anticipate and accommodate theuse of administrative data. TheCensus Bureau should work withorganizations representing thosetypes of GQs that were the mainsources of administrative data inCensus 2000 to evaluate how bestto gain complete information foreach GQ resident in an environ-ment where heavy use of adminis-trative data may be unavoidable.

Bring the SPOS on board earlier inthe schedule and maintain thisposition throughout nonresponsefollowup to oversee any SP/GQquestions/inquiries.

Develop a unique process to hirepersons without conventionalhousing to assist in the SBE opera-tions.

Train special place staff on how todeal with the public at all levelsfrom leaders of organizations tothe respondents.

Design the training schedule toallow time for the enumerators toabsorb what they have learnedbefore using the procedures in thefield.

Emphasize and improve thegeocoding and map spotting train-ing through all the operations.

U.S. Census Bureau Special Place/Group Quarters Enumeration 17

18 Special Place/Group Quarters Enumeration U.S. Census Bureau

Improve the special place informa-tion in other field operations mate-rials so there is a consistent under-standing and approach to specialplaces in all field operations.

Reduce the number of contacts tofacilities to prevent ill will beforeenumeration begins.

Begin GQ enumeration earlier sothat closeout does not competewith nonresponse followup.

Research and develop a method tolink a GQ questionnaire with a spe-cific GQ other than the currentprocess of a handwritten ID oneach questionnaire. The researchmust consider the data capturetechnology requirements in con-junction with the requirements offield automation.

Continue research on methods forenumerating people without con-ventional housing at service loca-tions. The SBE operation appearsto be a successful method ofincluding people without conven-tional housing in the census.

Continue research on methods forenumerating people at targetednon-sheltered outdoor locations.

Remove T-Night Enumeration fromSP/GQ Enumeration operations.One option is to use Update/Enumerate5 operations for T-Nightlocations.

Devise a method to link the HUquestionnaires filled out at T-Nightlocations with the T-Night loca-tions. This can be used to deter-mine counts of the population enu-merated at each T-Night location.

Reduce the size of the QA reinter-view sample by relaxing the popu-

lation estimate check’s tolerancelevel to reduce the number ofsmall GQs included in the sample.

Modify enumeration procedures forcertain types of GQs since someGQ types (e.g., Colleges andUniversities and Nursing Homes)required more rework based on theQA reinterview. Schoch (2003)also suggests reevaluating the vari-ables/criteria that select a GQ forreinterview and limiting the possi-ble cases for reinterview to certaintypes of GQs.

4.3 Data capture andprocessing

4.3.1 Usual Homes Elsewhere(UHE)

As reported in Jonas (2003), theprocess designed to deal with allCensus forms without a MAF ID,including GQ person records claim-ing a UHE, was called the Non-IDProcess.

Not all GQ records providing theaddress of a claimed UHE weresupposed to go into the Non-IDProcess. Records with a UHEaddress were to be screened forexclusion:

• By GQ type

• By the outcome of a screeningquestion on each GQ question-naire

The screening by GQ type wasdone after the initial Non-ID pro-cessing, returning 1,892,742

records to their original GQs.Excluding UHEs from the Non-IDProcess for certain GQ types wasdone to prevent people in certainUHE-ineligible types of GQs frombeing improperly enumerated at aresidence other than their GQ.

GQ questionnaires also were sup-posed to be screened from inclu-sion in the Non-ID Process by theirresponses to the residence ques-tion on each type of questionnaire.(For example, “Do you live or stayhere most of the time?” on theICR.) This screening was intendedto ensure that if persons whoseprimary residence was the GQ alsoprovided a UHE address, theywould not be enumerated else-where on that basis. Excludingcases from the Non-ID Processbased on the residence screeningquestions never took place. Of the1,048,536 records that underwentthe full Non-ID process, 388,970would have been excluded if theresidence screening question hadbeen used as intended. Becausethe procedures to screen GQ ques-tionnaires out of the Non-IDProcess were applied incorrectly,37 percent of the GQ question-naires ultimately resolved by theNon-ID process were in thatprocess inappropriately.

Of the 1,048,536 GQ personrecords that were ultimately includ-ed in the Non-ID Process:

• About 55 percent were matchedto an HU.

5 Update/Enumerate is a method of datacollection in which enumerators canvassassignment areas to update the address list-ing pages and maps and to conduct inter-views.

Table 3.Non-ID Outcomes

MilitaryUHEs

dropped

Matchedto a GQ

orreturnedto a GQ

Matchedto an HU

Total

Number Percent

Incorrectly . . . . . . . . . . . . . . . . . . . . . 3.0 22.8 11.4 388,970 37.1Correctly . . . . . . . . . . . . . . . . . . . . . . 17.9 1.6 43.4 659,566 62.9Total . . . . . . . . . . . . . . . . . . . . . . . . . . 20.9 24.3 54.8 1,048,536 100.0

Source: Jonas (2003) Table 8.2b.

• About 24 percent were eitherreturned to the GQ from whichthey came or were matched toanother GQ.

• About 21 percent were MilitaryUHEs that did not geocode andwere dropped from the Census.Most of these (about 86 percentof those dropped) were correctlyincluded in the Non-ID processand were dropped to avoidpotential duplication of militarypersonnel.6 The remaining 14percent (3 percent of the totalNon-ID process) should not havebeen included in the Non-IDprocess and were lost to theCensus.

• In addition to the 3 percent mili-tary UHEs that were dropped,another 11 percent thatmatched to HUs but that shouldnot have been included in theNon-ID process were lost fromthe GQ universe. These wereincluded in the Census in theHU universe.

4.3.2 Household forms and becounted forms included in GQ enu-meration

Jonas (2003) also reported on thenumber of persons enumerated onhousehold and Be Counted formsthat were included in the GQ uni-verse.

Some addresses had been in boththe HU universe and the GQ uni-verse in early versions of theDecennial MAF. These duplicateswere identified after the Censusmailing list was compiled. Thus,these units were mailed a Censushousehold questionnaire. If the HUquestionnaire was returned for aGQ, the persons enumerated on

the HU form were included in thecount for the GQ at that address.

The Be Counted program providedan opportunity for people withoutconventional housing to completea questionnaire if they had reasonto believe they were not enumerat-ed through other SP/GQ opera-tions. People who completed a BeCounted form and indicated thatthey had no address by markingthe “No Address on April 1, 2000”box on the form or indicated in theaddress section that they did nothave an address were included inthe SBE universe.

Table 4 indicates the numbers ofpersons who were added to GQsfrom household and Be Countedforms:

• Approximately 6.2 percent ofthe Group Homes populationcame from Be Counted formsand household questionnaires.

• Approximately 5.6 percent ofthe population in the categorySBE facilities and other GQscame from Be Counted formsand household questionnaires.

• For each of the other GQ cate-gories, no more than 0.4 per-cent of the population camefrom Be Counted forms andhousehold questionnaires.

4.3.3 Duplication in the GQ universe

4.3.3.1 SBE duplication

Since the SBE operation was con-ducted over a three day period, itwas possible to enumerate peoplemore than once. For example, ifsomeone used a shelter on March27 and received services at one ormore soup kitchens and/or regu-larly scheduled mobile food vanson March 28 they may have beenenumerated at each of these serv-ices. Also it was possible for per-sons who received services to fillout a Be Counted form. Anattempt was made to unduplicatethe SBE enumerations and counteach person only once in the cen-sus.

As part of the unduplicationprocess, individual demographiccharacteristics were assigned aweight based on whether theyagreed or disagreed. Agreementweights had positive values anddisagreement weights had negativevalues. Variables that were miss-ing from one of the two personrecords involved in the comparisonwere assigned a weight of zero. Afinal weight assigned to the pair ofperson records was the sum of theagreement and disagreementweights for each matching charac-teristic. Two or more person

U.S. Census Bureau Special Place/Group Quarters Enumeration 19

6 The decision to not return ungeocod-ed military UHEs was based on 1990 CensusSearch/Match results that showed peopleclaiming a UHE on MCRs and SCRs were list-ed at the HU address.

Table 4.Persons in GQs from Household and Be Counted Forms

Group quarters category Be countedforms HU records Total Percent*

1: Correctional facilities. . . . . . . . . . . . . . 30 136 166 0.02: Juvenile institutions . . . . . . . . . . . . . . . 1 494 495 0.43: Nursing homes. . . . . . . . . . . . . . . . . . . 1,925 3,902 5,827 0.34: Hospitals . . . . . . . . . . . . . . . . . . . . . . . . 140 681 821 0.35: Colleges and universities . . . . . . . . . . 514 3,011 3,525 0.26: Military installations . . . . . . . . . . . . . . . 10 11 21 0.07: SBE facilities and other GQs . . . . . . 33,264 14,246 47,510 5.68: Group homes . . . . . . . . . . . . . . . . . . . . 724 27,979 28,703 6.2

Totals . . . . . . . . . . . . . . . . . . . . . . . . . 36,608 50,460 87,068 1.1

*Percent of GQ population.

Source: Jonas (2003) Table 8.3.

20 Special Place/Group Quarters Enumeration U.S. Census Bureau

records were considered a match ifthe cumulative weight was above0.95. The record for the primarysource within each pair was includ-ed in the census. Refer to McNally(2002) for more details on the SBEunduplication process.

Table 5 provides the results of theunduplication of SBE personrecords at the national level asreported in McNally (2002). Thefirst row represents the total num-ber of data captured personrecords. The second row providesthe number of person records thatmatched to another census personrecord and were not counted in thecensus. The third row indicatesthe total number of unique (undu-plicated) people that were includedin the Census 2000 results. Thefourth row indicates the number ofperson records with sufficient datafor matching. The fifth row repre-sents the number of personrecords with insufficient data formatching.

A total of 241,941 data capturedpersons from SBE locations wereincluded in the Census. Of those,approximately 87 percent hadenough information for the ques-tionnaire to be included in theunduplication process.

McNally (2002) also noted thatnumerous records were erroneous-ly identified as duplicates. Of the16,787 person records matchedand unduplicated during data pro-cessing, 2,410 (14 percent) were most likely erro-neously unduplicated. Targetednon-sheltered outdoor locationshad the highest percentage (39 percent) of erro-neous duplicates.

4.3.3.2 Within-GQ person duplication

Jonas (2003) conducted an evalua-tion of within-GQ person duplica-

tion based on early non-systematicobservations of Census data thatsuggested there were a significantnumber of duplicate personrecords within GQs, particularly ingroup homes and other small GQs.He designed a stratified sample of400 GQs in five strata to estimatethe magnitude of duplication with-in the GQ population. It excludedcorrectional facilities, militaryinstallations, and SBE facilities7, butincluded the rest of the GQ uni-verse. The portion of the GQ uni-verse from which the sample wasdrawn included 154,042 GQs con-taining 5,156,168 person records,or 66 percent of the GQ popula-tion. The 400 GQs in the samplecontained 18,650 person records.

The person records in each GQwere clerically examined to identi-fy duplicates. Records with thesame name, sex, and age/date ofbirth were considered duplicates.The estimated number8 of dupli-cate person records is 56,416 ±34,409, which is 1.1 percent ± 0.7percent of the persons in GQs fromwhich the sample was drawn.

Group homes and religious GQswere found to be the largest singlesource of duplication, apparentlybecause many such facilitiesreturned household questionnairesin addition to being counted by GQenumerators.

4.3.3.3 People in HUs duplicated to people in GQs

Feldpausch (2001) reported on thenumber and percent of E-samplepeople (i.e., people enumerated incensus HUs in the A.C.E. sampleblock clusters) duplicated to peo-ple outside the search area. Asshown in Table 5, she found therewere over a half million E-samplepeople duplicated outside thesearch area to people in groupquarters, with over half the dupli-cates to college dorms. Othertypes of GQs with a high percent-age of duplication based onFeldpausch’s results were nursinghomes (about 10.9 percent), localjails (about 8.5 percent), militarybarracks (about 4.5 percent), andstate prisons (about 4.5 percent).9

A person found duplicated wascaptured twice by the Census. Therecord of the person that was

Table 5.Results of the SBE Unduplication of Data Captured Person Records

Number Percent

TOTAL data captured person records from SBElocations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258,728 100.0

Data captured person records matched and notcounted. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16,787 6.5

Data captured person records counted in theCensus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241,941 93.5➢Data captured person records with sufficient

data for unduplication . . . . . . . . . . . . . . . . . . . . . . . . . 209,488 86.6➢Data captured person records with

insufficient data for unduplication . . . . . . . . . . . . . . . 32,453 13.4

Source: McNally (2002).

7 SBE facilities were excluded becauseunduplication was already being done on theservice-based population; correctional facili-ties and military GQs were excluded becauseof the unlikelihood of matching HUs to pris-ons and barracks on the Master Address File.

8 Estimates are with a 95 percent confi-dence interval.

9 The estimate of duplication betweenhousing units and group quarters in theA.C.E. Revision II was similar to the ESCAP IIestimate. There are no additional estimatesfrom A.C.E. Revision II on people duplicatedto GQs outside the A.C.E. search area.

captured in the correct placeaccording to census residencerules should have been coded as acorrect enumeration. The recordof the person that was captured inan incorrect place according tocensus residence rules should havebeen coded as an erroneous enu-meration.

Assuming the GQ enumeration wasperfect, those people duplicated toGQs that could not claim a UHEshould all be considered erroneousenumerations. For the majority ofthese people, it seems likely thatthe GQ was their April 1 residence.

Assuming the GQ enumeration wasperfect, those people duplicated toGQs that could claim a UHE shouldall be considered correct enumera-tions. For most of these people, itseems likely that the HU was theirusual April 1 residence.

4.3.4 Recommendations

The following recommendationswere made by Feldpausch (2001),Jonas (2003), McNally (2002), andSchoch (2003a):

Institute safeguards to make itmore likely that if the same specialplace or GQ is deleted and re-added, it is identified as the sameentity and identified by the sameCensus identification number.

Create and maintain a source codeand history of all actions for eachfacility from beginning of initial filecreation through processing.

Track individual GQ questionnairesthrough post-enumeration process-ing, from enumeration throughdata capture. In Census 2000,each GQ questionnaire had aunique barcoded number printed

on it; however, the barcode wasnot used to track GQ question-naires.

Institute more effective softwarequality assurance programs.

Reevaluate the definitions to clarifythe distinctions between what con-stitutes a GQ versus an HU. Aspointed out by Jonas (2003), thelargest single source of duplicationwas a result of small group homesbeing enumerated as housing unitsas well as GQs.

Research ways to improve the SBEunduplication process.

Research ways to revise residencerules instructions to reduce dupli-cation between HUs and GQs.Some possible explanations forthis duplication include:

• The instructions indicating whoto include may have been mis-understood or ignored by therespondent. (Examples of peo-ple enumerated in HUs in errorinclude college students, peoplein local jails, and people in nurs-ing homes.)

• The respondent may not haverealized that a household mem-ber was enumerated elsewhere.

• Some GQs were enumerated

using administrative records

that did not reflect the residents

as of April 1, 2000.

There were many specific recom-

mendations included in the evalua-

tions, QA profiles, operational

assessments, and headquarters

debriefing report. Many of those

are included in earlier sections of

this report. This section discusses

the major recurring themes that

appeared across the source docu-

ments.10 As noted below, the

SP/GQ 2010 Research and

Development (R&D) Planning Group

is already working on classification

issues, SP/GQ inventory develop-

ment operations, and integrating

the HU and GQ address lists and

operations. Research and develop-

ment for GQ enumeration and pro-

cessing is also planned.

U.S. Census Bureau Special Place/Group Quarters Enumeration 21

Table 6.E-Sample Duplicates to People Outside the A.C.E. Search Area inGQs (Standard Errors)

Could not claim a UHECould claim a

UHENot a dorm Dorm

Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189,756 271,158 59,586(11,028) (34,806) (5,915)

The numbers are weighted with the final E-Sample weight. They only include caseswhere the model probability of being a duplicate is greater than 0.5.

Standard errors are shown in parentheses.

Source: Feldpausch (2001) Table 3.

10 Source materials suggest that therewere a sizable number of problems withgeocoding special places and GQs and thatthese problems resulted in geographic misal-location of GQs in data tabulations.Reviewers of this paper suggested geocod-ing problems as a recurring theme.However, while the Census Bureau is awareof the geocoding problems and is working toaddress them for the 2010 Census, there isno formal documentation that could be usedas input for this report.

This page intentionally left blank.

5. Synthesis of Results andRecommendations for the 2010 Census

U.S. Census Bureau Special Place/Group Quarters Enumeration 23

5.1 Revise group quartersdefinitions andclassification of groupquarters type codes

According to Schoch (2003a), theCensus 2000 SP/GQ definitionsand classifications were the foun-dation of the SP/GQ inventorydevelopment process. These defi-nitions and classifications weredeveloped to provide the parame-ters for determining the SP/GQ uni-verse and to further define thespecific type of facility.

Although the definitions and classi-fications were researched and test-ed over the decade, there wereinherent weaknesses in the overallidentification and classification ofsome facilities, which ultimatelyimpacted the final enumeration.These were:

• The definitions and the termi-nology used by the CensusBureau were not always consis-tent and/or current with that ofother agencies, state, local, andtribal officials, and the generalpublic that resulted in discrep-ancies within the final SP/GQuniverse.

• The concept of embedded HUsin GQs (such as a resident advi-sor’s apartment in a college dor-mitory) was not clearly definedand was generally misunder-stood by both staff and the gen-eral public. The lack of clarityassociated with this conceptimpacted not only the identifica-tion of such units, but con-tributed to the duplicationbetween the GQ and HU inventories.

• There were also duplicatesbetween the GQ and HU inven-tories as a result of the failureto match the two separateinventory files and flag dupli-cate addresses. Another factorcontributing to duplication wasthe lack of clarity and consisten-cy in defining certain types ofGQs.

Recommendations fall into twobroad categories.

• Revise the GQ definitions sothat they are current, agreewith terminology used byfacility service providers,advocates, government offi-cials, and the general public,and can be easily under-stood and implemented byfield staff. As part of the les-sons learned from the Census2000 Group Quarters operationsand the 2010 Census planningeffort, an interdivisional team ofCensus Bureau group quartersexperts was formed to researchand revise the Census 2000group quarters definitions anddetailed classifications. Furtherresearch will be undertaken todiscuss these definitions withindustry experts, representa-tives from local governments,and data users. Plans are toconduct focus groups and con-ferences to discuss the defini-tions, conduct cognitive testingto assess the changes, and con-duct field work to test andimplement the definitions.

• Clarify the distinctionsbetween GQs and HUs. Manyof the smaller GQs are indistin-

guishable in appearance fromsingle-family residences. Thesefacilities, therefore, may havereceived an HU questionnaire inthe mail from the USPS or dur-ing the Update/Leave11 opera-tion and also may have beenvisited during GQ enumeration.One option that is being pur-sued by the SP/GQ 2010 R&DPlanning Group is to enumeratethese as HUs but tabulate theirpopulation as part of the GQuniverse.

Include evaluation of the clas-sification of GQs based on therevised definitions in futureresearch. It was anticipated thatStevens (2003) would provideempirical results on classificationissues; however, because of thedesign limitations of the study(evaluating special place ratherthan GQ type codes), the resultsare not pertinent for planning the2010 Census.

5.2 Integrate housing unitand group quartersaddress list developmentactivities

Re-evaluate and design amethod to integrate the SP/GQinventory development into theoverall MAF developmentprocess. Develop an early inte-gration and reconciliation of thegroup quarters and housing unitfiles. As previously stated, the

11 Update/Leave is a method of data collection in which enumerators canvassedassignment areas to deliver a census ques-tionnaire to each HU and update the addresslisting pages and maps. The household wasasked to complete and return the question-naire by mail.

duplication across inventoriesexisted primarily due to the failureto match the two separate invento-ry files.

Integrate GQs with HUs in theLUCA operations. Because theSpecial Place Facility Questionnaireoperation was not completed ontime, SP/GQ could not be loadedinto the MAF in time to be includedin the LUCA 98/99 programs andthis necessitated the separate SPLUCA program. Because of thisdelayed implementation, otheroperations were scheduled at thetime the SP LUCA was conductedand therefore it was difficult tomeasure the success of the individ-ual SP/GQ operations.

In order to resolve these issues,the SP/GQ 2010 R&D PlanningGroup, working with other plan-ning and implementation teams, isdeveloping and testing new proce-dures to integrate the HU and GQlisting operations. One part of thiseffort is to ensure that addresslists for all field operations includeboth HUs and GQs. Instead oftraining all listers on identifyingand classifying GQs, the listersmerely have to identify units forwhich they are unsure of the statusas “other living quarters”. Trainedenumerators will go out to theother living quarters and using aspecially designed Other LivingQuarters Questionnaire, determineif the unit is an HU or, if a GQ,classify the GQ by the appropriatetype code.

5.3 Begin planning thegroup quartersenumeration early

Begin planning SP/GQEnumeration operations earlyin the decade so all plans arecompleted and in place to allow forconsistency and timely completionof all the associated materials.

Ideally, all planning, procedures,and materials for the 2010 Censuswould be complete in time for thedress rehearsal in 2008.

Early planning for the 2010 Censusenumeration of GQs has alreadybegun. The first meeting of theSP/GQ 2010 R&D Planning Grouptook place in December 2001.Most of this early work centers onSP/GQ inventory developmentissues. A new procedure, GroupQuarters Validation, using theOther Living QuartersQuestionnaire will be tested in the2004 Census Test.

Plan more tests involvingSP/GQs to make sure we haveused every opportunity to test allparts of the operation from SP/GQinventory development to enumer-ation of the GQs. In addition totesting address list developmentprocedures in the 2004 CensusTest, other research and develop-ment work is planned, includingfocus groups and meetings withindustry experts and serviceproviders and special purposetests to address GQ issues thatmay not be suitable for a site test.This work will be further devel-oped for the 2006 Census Test.

5.4 Continue research onservice-based facilityenumeration operations

Continue supporting the SBEoperations for the 2010Census. According to McNally(2002), the service-based enumera-tion operation appears to be a suc-cessful method of including peoplewithout conventional housing inthe census.

Revise the timing of SBE opera-tions. Many of the operationalreports indicated issues with tim-ing of the SBE operations. Some ofthe timing issues concern the over-lap of the SBE operations, training,

and T-Night operations and com-plaints from soup kitchens thatwere not open on the one day des-ignated for enumeration.

There are also data quality issuesthat are affected by the timing ofoperations. There is the potentialfor duplication when SBE opera-tions are scheduled for more thanone night or are conducted muchearlier than Census Day or earlierthan the other GQ enumerationoperations. These issues need tobe taken into account when deter-mining the schedule for opera-tions.

5.5 Track group quarters facilities andquestionnaires throughoutthe census

Jonas (2003) cited two global limi-tations that affected his evaluation.These were the absence of a sys-tem for tracking individual ques-tionnaires through the enumera-tion process and the limitations ofthe system designed to track spe-cial places and GQs from thebeginning of the Special PlaceFacility Questionnaire processthrough tabulation. These limita-tions could also affect the flow ofoperations and duplication of GQsin the Census.

Jonas (2003) offered two recom-mendations concerning these limi-tations:

• Maintain all special placeand GQ records throughoutthe census. A file that usesflags to indicate deletes, ratherthan permanently removingrecords from the file, would pro-vide much more complete infor-mation about the SP/GQ uni-verse over time. Institutesafeguards to make it more like-ly that if the same special placeor GQ is deleted and re-added, itis identified as the same entity

24 Special Place/Group Quarters Enumeration U.S. Census Bureau

and identified by the same cen-sus ID numbers.

• Track individual forms fromenumeration through datacapture. Differing counts ofthe number of completed ques-tionnaires were obtained formany GQs because the numberof questionnaires from each GQwas tracked, rather than thequestionnaires themselves. Theresult would be a much moreexact and reliable GQ popula-tion count.

5.6 Continue research onunduplicating people ingroup quarters

Review the residence rules toemphasize the instructions for

who to include in the house-hold. Based on some of the possi-ble explanations for duplicationmentioned by Feldpausch (2001)(e.g., respondents not understand-ing or ignoring instructions forwho to include/exclude, respon-dents including people who wereactually enumerated elsewhere,and some GQs being enumeratedusing administrative records thatdid not reflect the residents as ofApril 1), some duplication may bereduced or eliminated by changesto or emphasis on adherence tothe residence rules.

Reduce duplication betweenHUs and GQs. Jonas (2003)reported approximately 1.1 per-cent duplicate person records in

the GQ universe studied, largely aresult of small GQs returning ahousehold questionnaire and alsobeing counted during GQ enumera-tion. He suggested that reducingthe duplication between the HUand GQ universes would helpreduce within-GQ person duplica-tion.

Research ways to improve theSBE unduplication process. Theaccuracy of the SBE unduplicationwas affected by person recordswith insufficient data (13 percentof data captured persons records,while counted in the census, hadinsufficient data for unduplication)and by enumerators enteringsomething other than a name,such as “Client” in the name field.

U.S. Census Bureau Special Place/Group Quarters Enumeration 25

This page intentionally left blank.

There were approximately 7.8 mil-

lion people tabulated in GQs in

Census 2000. This is less than 3

percent of the total U.S. population

on April 1, 2000. In order to enu-

merate people in the GQ universe,

there are a variety of unique proce-

dures required to identify GQs,

assign them to the correct geogra-

phy, enumerate the residents, and

data capture and tabulate the per-son records.

While the Census 2000 enumera-tion of GQs appears to have beenoperationally successful, the evalu-ations and operational assess-ments identified various issuesresulting from list development,enumeration, and processing. Theevaluations and assessments also

provide recommendations for

improvements for the 2010 Census

GQ enumeration.

Although the Census Bureau has

already initiated research and test-

ing to address these issues and

recommendations, there is much

work that needs to be done before

the 2010 Census.

U.S. Census Bureau Special Place/Group Quarters Enumeration 27

6. Conclusion

This page intentionally left blank.

References

Feldpausch, Roxanne (2001),“Executive Steering Committee onAccuracy and Coverage EvaluationPolicy II Report Number 6: CensusPerson Duplication and theCorresponding A.C.E. Status”,October 2001.

Gerber, Eleanor and Tracey R.Wellens (1994), “CognitiveEvaluation of the Service BasedEnumeration (SBE) Questionnaire: ATale of Two Cities (and One MobileFood Van),” September 1994.

Gerber, Eleanor (1997), “FinalReport of the SBE New York SmallScale Test,” March 1997.

Gloster, John (2000), “ProgramMaster Plan: Census 2000 Service-Based Enumeration,” Census 2000Informational MemorandumNumber 40, February 2000.

Griffin, Richard A. (1993),“Sampling the HomelessPopulation,” February 1993.

Griffin, Richard A. (2001), “Census2000 - Service Based EnumerationMultiplicity Estimation,” DSSDCensus 2000 Procedures andOperations Memorandum Series B-15, February 2001.

Halterman, Kathleen (2000),“Census 2000 Local Update ofCensus Addresses Special PlaceProgram Master Plan,” Census2000 Informational MemorandumNumber 69, August 2000.

Johanson, Carrie and BroderickOliver (2003), “Profile of theMilitary Group Quarters AddressListing Quality AssuranceOperation,” Census 2000 QAProfile, February 2003.

Jonas, Kimball (2003), “GroupQuarters Enumeration,” Census2000 Evaluation E.5- Revised,August 2003.

Kalton, Graham, Daniel Levine,David Marker, and Laura Sharp(1994), “Methods to EnumeratePersons With No Usual ResidenceUsing Sampling and Estimation,”January 1994.

Long, John F. (1999), “ProcessingGroup Quarters with a Usual HomeElsewhere,” Internal Census BureauMemorandum from Long toMichael J. Longini, November1999.

Martin, Elizabeth (1995),“Evaluation of the OperationalProcedures for the Service-BasedEnumeration in the 1995 CensusTest,” DMD 1995 Census TestResults Memorandum Series No.13, November 1995.

McNally, Tracey (1999), “Service-Based Enumeration Coverage YieldResults,” Census 2000 DressRehearsal Evaluation MemorandumD1, April 1999.

McNally, Tracey (2002), “Service-Based Enumeration,” Census 2000Evaluation E.6, December 2002.

Medina, Karen (2003), “AssessmentReport for 1998, 1999, and SpecialPlace Local Update of CensusAddresses and New ConstructionPrograms, Census 2000Informational MemorandumNumber 140, July 2003.

Miskura, Susan M. (2000),“Comprehensive Census 2000Assessments,” DecennialDirectorate 2010 Census PlanningMemorandum Series No. 3,December 2000.

Mule, Thomas (2001), “ExecutiveSteering Committee on Accuracyand Coverage Evaluation Policy IIReport Number 20: PersonDuplication in Census 2000,”October 2001.

Oliver, Broderick (2002), “Profile ofthe Census 2000 Group QuartersReinterview Program,” Census2000 QA Profile, December 2002.

Schoch, Sharon (2001), “Census2000 Special Place /GroupQuarters Inventory DevelopmentProgram Master Plan,” Census2000 Informational MemorandumNumber 113, September 2001.

Schoch, Sharon (2002), “Census2000 Military InstallationEnumeration Program Master Plan,”Census 2000 InformationalMemorandum Number 117,January 2002.

Schoch, Sharon (2003a), “SpecialPlace /Group Quarters (SP/GQ)Inventory DevelopmentAssessment Report,” Census 2000Informational MemorandumNumber 131, January 2003.

Schoch, Sharon (2003b), “SpecialPlace/Group Quarters (SP/GQ)Enumeration OperationsAssessment Report,” Census 2000Informational MemorandumNumber 139, July 2003.

Schwede, Laurie (1994), “TheKentucky Count Overview andObservation Report,” July, 1994.

Smiley, Richard A. (1997), “AnEvaluation of the Quality of theDemographic Data Collected byEnumerators in a Test of Service-Based Enumeration Procedures,”Proceedings of the Survey ResearchMethods Sections, Joint StatisticalMeetings, August 1997.

Smith, Annetta Clark (1995), “1995Census Test Service-BasedEnumeration Operation TalkingPoints,” undated.

Smith, Annetta Clark (2003),“Revised Definitions for SelectedTypes of Group Quarters to beTested in the 2004 Census Test,”briefing for the National Academyof Sciences, March 2003.

U.S. Census Bureau Special Place/Group Quarters Enumeration 29

Smith, Robert Ellis (2003), “Reportto the Bureau of the Census onPossible Restrictions to Access byCensus Workers to Certain “GroupQuarters”,” Privacy Journal, June2003.

Stackhouse, Hub (2000), “GroupQuarters Enumeration ProgramMaster Plan,” Census 2000Informational MemorandumNumber 41, February 2000.

Stevens, Neala (2003), “Evaluationof the Facility Questionnaire(Computer Assisted TelephoneInterviewing and Personal Visit),”Census 2000 Evaluation E.1b, May2003.

U.S. Bureau of the Census (1993),1990 Census of Population andHousing History, 1990 CPH-R-2,Part A, October 1993.

U.S. Bureau of the Census (1995),1990 Census of Population andHousing History, 1990 CPH-R-2,Part C, October 1995.

Walker, Shelley (2001),“Questionnaire Assistance Centerand Be Counted Program MasterPlan,” Census InformationalMemorandum Number 103, May2001.

Zajac, Kevin J. (2003), “Analysis ofImputation Rates for the 100Percent Person and Housing UnitData Items from Census 2000,”Census 2000 Evaluation B.1.a, July2003.

30 Special Place/Group Quarters Enumeration U.S. Census Bureau

Census 2000 Topic Report No. 5TR-5

Issued February 2004

SpecialPlace/GroupQuartersEnumeration

U.S.Department of CommerceEconomics and Statistics Administration

U.S. CENSUS BUREAU

Census 2000 Testing, Experimentation, and Evaluation Program