Decennial Census: Preliminary 1990 Lessons Learned Indicate Need to Rethink Census Approach

Published by the Government Accountability Office on 1990-08-08.

Below is a raw (and likely hideous) rendition of the original report. (PDF)

                 United-States   General   Accounting   Office   / 4    4651



For Release        Decennial Census: Preliminary 1990 Lessons Learned
on Delivery        Indicate Need to RethinkCensus Approach
Expected at
lo:30 a.m. PUT
August 8, 1990

                   L. Nye Stevens, Director
                   Governmznt Business Operations Issues
                   General Government Division
                   Before the
                   Subcuixnittee on Census and Population
                   Ckttee      on Post Office and Civil Service
                   House of Representatives


          --                                                           GAO Form 160 (12/87)

                                    SUMMARYOF STATEMENT OF
                                        L. NYE STEVENS
                                 DIRECTOR, GOVERNMENTBUSINESS
                                       OPERATIONS ISSUES

         Escalating     census costs,    declining    public    cooperation,      and a
         shrinking     workforce   are among the challenges         that indicate     the
         need for a reassessment        of the 1990 decennial        census methodology.
         Key decisionmakers--Congress,           the Census.Bureau,       other executive
         agencies,     data users,    and statistical     experts--must       begin to
         rethink    the current    census methodology       while the experience        of
         the 1990 census remains fresh and opportunities                for change exist.
         These decisionmakers       need to begin a dialogue         now on several
         important     questions.
         First   and foremost,        agreement    is needed on whether        the decennial
         census should be streamlined             and alternative       data collection
         approaches      developed     ,for some important       socioeconomic     data.
         Second, alternatives           to the 1990 census methodology           need to be
         explored.       A critical      question    for future     census methodologies
         focuses on the extent            to which sampling can be used.           Dialogue  on
         these questions        will   serve as a catalyst        to examine the Nation's
         larger    information       needs and how to best meet them.
         One vehicle     for focusing     discussions       on these issues would be a
         commission     or panel such as the one proposed             in the Federal
         Information     Resources Management Act (S. 1742).                 In February
         1990, GAO testified        in support     of this bill's       provision    to create
         a commission      that would identify        long-term    information      needs and
         set priorities       for information      in light     of changes in the economy
         and society.        GAO believes     such a commission       could also assess
         alternative     census methodologies.
Mr.     Chairman             and Members of                   the      Subcommittee:

We welcome              this         opportunity              to contribute                    to your              hearing         on the
preliminary              lessons             learned          from         the      1990 census.                     We believe
that      holding             this       hearing           as part          of the         annual              meeting         of     the
American          Statistical                  Association                 is particularly                      appropriate
because          &t underscores                      the     central             role     of        the decennial                census
in providing                 vital          statistical             information                     for      the     ensuing

We have been monitoring                               the     1990 census                 since             the mid-1980s              and
have testified                   before          this        Subcommittee                 on numerous                  occasions            on
the     status          of     the       census,           operational                  problems             encountered              in
different           census            operations,                and actions               taken             by the Bureau                 in
response          to these               problems.             Although                 many important                  census
operations              are      still         underway           or pending,                  it         is not       too     early        to
begin       to examine                the      implications                 of     the     1990 census                  experience
for     the way we carry                       out      future         decennial               censuses.

Essentially,                 we believe               that     a series                 of challenges                  to taking            the
census,          some        of which           are        longstanding,                  indicate              the     need for
rethinking              the census              methodology.                      Besides            its      use      as a forum
for     discussions                  about      rethinking                 the     census            methodology,               we
believe          that        today's          hearing            is a critical                      first       step         toward
creating          an environment                      to begin             discussions                    on the       Nation's
information              needs           and how best               to meet them.                           While      the     decennial

census         was never               intended         to meet all                the    Nation's              information
needs,         it      is clearly              the    cornerstone                of    the          Nation's          statistical
effort.              We live           in an emerging               information                     age with          an
increasingly                 integrated              world        economy;            public          and private                 sector
decisionmakers                   require         timely,           complete,             and accurate                  information
to successfully                   manage and compete                        in this          dynamic            environment.

Reassessing                the    current            census        methodology                 is clearly              a major
part      of        the    agenda        for     discussing               the Nation's                  information                and
statistical                needs.          Efforts           to    reexamine             the         census      methodology
and our             national        statistical               policy            need to begin                  in earnest
because         key decisions                   need to be made soon                           if     new data
collection                strategies            are    to be tested                   in time           for     the        next
decennial             census.


The basic             census        methodology               is essentially                        a "head      count"
whereby             the Bureau           attempts            to take            a "national              snapshot"                as of
April      1-0 counting                every         individual            one by one,                 household             by
household,                in every        community.                In addition                     to this      head count,
the decennial                  census          is used to collect                      information               on the
nation's             housing        stock        and a wide               variety         of other
socioeconomic                  data.           For example,              the      1990 decennial                  census
short      form           questionnaire               asks questions                   on the          type      of housing
ur@t,      the        number of           rooms in the                  unit,         and the market                   value        of

that         unit.           In addition                   to these         questions,        the      long      form        also
asks         such questions                     as the          number      of automobiles,             vans,           or
trucks          that        are kept             at the housing                unit     and the        type      of      fuel
that         is used         for         heating.

For     the          1990 census                the      Bureau        asked    about      95 percent            of      the
Nation's              estimated             106 million                households        to mail        back       their
census          forms;         the        other          5 percent          were asked        to hold           their
questionnaires                     for      pick-up             by census       enumerators.               The Bureau's
procedures              called            for      it      to make up to six              contacts         per        household
that     did         not    mail          back          their     questionnaires.                 If    after           repeated
attempts              the   Bureau              could       not    obtain       population          data      directly
from     the household,                         the Bureau             would    generally         obtain         data        from
other         sources          such as neighbors,                        postal       carriers,        or building

We believe              that        several              challenges,           some of which           are
longstanding,                  have affected                     the    1990 census         and may become even
more difficult                     to address               in the       future.         Among these             are
escalating              census            costs,          a declining           level     of public           cooperation,
and a shrinking                     workforce               available          for    temporary        employment.
We believe              these            challenges              indicate       the     need for       a reassessment
of     the      1990 census                methodology.
Escalating            costs

Decennial           census           costs             have risen                 dramatically                 in recent             decades.
The Bureau            spent           (in        1980 dollars)                     $5.04         in     1950 and $12.10                    in
1980 to count                each housing                        unit.            In 1990,            the     cost        to count           each
housing        unit      will          increase                  again,           and,      based        on current                 Bureau
estimates,            will          rise      to about                   $17.00          (in    1980 dollars).                       Overall,
the Bureau            estimates                  it     will        spend about                 $2.6         billion          for     the
1990 census.                  This          is a significant                         increase            over          the    $1.1
billion        spent          for      the            1980 census                 and $221 million                       spent       for     the
1970 census.                  In terms                 of constant                 1980 dollars,                   the       cost     of     the
1990 census            will          have             increased            about          73 percent               over       1980 costs.
If   the    Bureau           uses          the         1990 methodology                        for     the      2000 census,                 the
Bureau's        21st         century              planning                staff          estimates            that        the       2000
census      could        cost          $4.8            billion            (in      current            dollars)'.              These costs
assume a 55-percent                          mail         response                rate      and a lo-percent                        increase
in the      workload.

Declining           Public           Cooperation

Reduced public                 cooperation,                       another           continuing                challenge,              has
significant            cost          and data               quality               implications.                    The mail
response        rate         for       the            1990 census               was 7 percentage                        points        below
what      the Bureau                expected              and 12 points                     below        the       rate       obtained
during        the     1980 census.                        Preliminary                    results         from        an evaluation
of &he Bureau‘s                     promotion               program               indicate            that      almost           93 percent

.                        .

        of   the     general            public          was aware             of the        census            just      prior        to
        Census       Day.         However,              awareness             alone        did     not        sufficiently
        motivate             a large       segment              of the        public        to         respond.          There           has
        been considerable                       speculation              in       the media             and from         the        Bureau
        about       the       reasons           for     the      lower-than-expected                          mail      response             rate.
        Potential             explanations                offered         include           a growing                reluctance              to
        respond          to questionnaires                      and surveys;                a perception                 that        the
        census       questionnaire                     is burdensome                and/or         intrusive;                and a
        general          increase          in the          volume         of mass mailings,                          which        may make
        it   difficult            for      the         public       to    identify               the     census         questionnaire.

        The lower-than-expected                            mail      response              rate         increased            the     cost         of
        follow-up            activities               by about           $70 million;                   the    increase             was
        primarily            used       to hire          additional                field     staff            to gather             census
        data.        An additional                     $14 million                was needed             to cover            the     costs           of
        a pay increase                  instituted              to expedite                the     completion                of
        nonresponse             followup.                As a result,                 nonresponse               followup             field
        costs       rose       to almost               $260 million.                  If    public            cooperation
        continues             to decline,               the     current            census         methodology                will        be even
        more costly             for       the         2000 decennial                census.

        Even after             incurring               the costs          for       personal             follow-up             visits,            the
        Bureau       still       had to collect                     surrogate              population                data;        that       is
        population             data      from         sources        other          than     the         household             itself.
        Some major             urban       areas         had a disproportionate                               percentage             of      their
        population             enumerated               with      surrogate            population               data.

Nationwide,           the         Bureau      collected                 surrogate          population              data       for      14
percent       of the         occupied              housing          units         it     visited          during
nonresponse           followup.               However,              in some urban                  areas,         the      Bureau
relied      on surrogate                  population            data        for         as many as 40 percent                         of
the      occupied       housing            units       it     visited             during         nonresponse
followup.            These statistics                       raise        questions           about         the     cost-
effectiveness               of     followup           procedures             and the quality                     of
population          data         obtained,            particularly                     in major      urban         areas.

Staffing       difficulties

Another       longstanding                 challenge            for        the     Bureau          has been recruiting
the      army of workers                  needed       to take             the     census--the              majority            of
whom are       required             for     nonresponse                  followup.               Despite         the
Bureau's       actions             to implement               geographic                 pay rates          and the
enactment          of legislation                   enabling             federal          annuitants             and
military          retirees          to work           on the            census         without       reductions               to
their      census      salary             or retirement                  benefits,           the     Bureau           still
experienced           widespread              staff          shortages.                  These shortages                   delayed
the     completion           of     the     1990 nonresponse                       follow-up          operation.

The Bureau          took         several           actions          that     generally              put     the       census
back on schedule.                    One such action                       was to         implement          a pay raise
during      the     final         stages           of nonresponse                  followup          to address               staff
shortages.            However,             staff       shortages             may be more severe                           in the
future      because          workforce              projections              for         2000 indicate                a
continuing           decline              in the      number of available,                       qualified          workers.


The fundamental                    constitutional              objective             of the decennial                 census
is obtaining               a complete             and accurate               population           count      for
reapportioning                    the     House of Representatives.                             However,      decennial
census       data         are      also      used to redraw                 state      and municipal
legislative               districts             and will       be used during                   the decade          for
allocating           billions              of    federal       and state             dollars.           In addition,
businesses           and researchers                      make extensive               use of decennial                   census
data,      including               socioeconomic              data.

While      socioeconomic                  data      are      important,             decisionmakers            in
Congress,           the         executive         branch,          and the private                sector      need to
consider       whether              the decennial              census          is    the    appropriate
mechanism           for         collecting         socioeconomic                dataL       particularly             given
the     increasing               demands for              timely       and accurate             data.        After
decisionmakers                   agree       on the        scope       of    the     information           needed          for
the     census,           alternatives              to the         1990 census             methodology             need to
be explored               for      taking        future       censuses.

First      and foremost,                  agreement           needs         to be reached            on whether             the
decennial         census            should        be streamlined                and whether             alternative
data     collection                approaches             should       be used for           gathering             important
socioeconomic                   data.        We believe            some of          the critical           questions             on

      the   scope        of information                     needed      from       the decennial             census
     include       the         following:

             What,        if      any,       short          form     housing       data       should       be collected                  on
               a sample           basis?
             What,        if     any,        of the          socioeconomic             data      currently           included
               on the          long      form        could     be deleted?
            What alternative                         sources         exist     or could          be developed                 to
             obtain            information              currently            collected          on the       long     form
             Could       alternative                  data     collection            approaches            for
             socioeconomic                   data       improve         data     quality         and timeliness?                    At
            what      cost?
            To what             extent         should         the     Bureau       collect,          tabulate,            and
            disseminate                 data         that     are primarily               used by the private
            Should         private             sector         data     users       share       the     financial              burden
            associated                with      collecting             certain         socioeconomic               data,           and
             if    so,     to what            extent?

Once agreement                     is     reached            on the        scope     of      information           that        should
be obtained                during            the decennial                 census,        alternatives             to the           1990
census            methodology                need to be explored.                         Experience         with         previous
censuses            has shown that                     even        the best        attempts          to ensure            a
complete            enumeration                 of     the population                still      result           in some
poStion            of the population                        being      missed,         particularly              certain

minority           population                groups.           Although             the      net       national         undercount
has declined                over        the        past     four      decades,              according           to the
Bureau's           estimates,                the disproportionality                              of    the undercount              has
not.        Missed       persons              may result              from         either          an incomplete
address        list      or missing                  persons         within          a household.                  In addition,
some segments                of    the population,                       such as undocumented                          residents
or households                in public               housing         projects             with         more than         the
number        of authorized                   residents,             may want               to avoid           being     counted.

Also,       as I stated                earlier,             even after              several            personal         visits      to
households            that        did        not     mail     back         their       questionnaire,                   the Bureau
had to obtain                surrogate               population              data      for         a substantial             number
of housing            units,           particularly                 in some urban                     areas.      Given          these
conditions,            several               questions             about      the      census             methodology
should       be considered.                         These     include          the        following:

--         How extensively                    can sampling                 be used           for       the decennial
           census,       considering                  legal         and data           quality             issues?         For
           example,          should           the Bureau             visit          a sample              of households            that
           do not      return            census           questionnaires                    in     lieu     of visiting
           every      nonresponding                   household?
--         What is       the best              method         to develop               and maintain               a complete
           address       list          for     mailing         questionnaires?
--         Could      the       U.S.         Postal         Service          be used more extensively
           during      the enumeration                       process?

In addition           to questions                 about      the      census          methodology,                another
important          area      of    inquiry          focuses          on the         causes           for      declining
levels      of public             cooperation,               which      is a key determinant                            of       the
accuracy       and completeness                     of     the census            counts.              Although             the
effectiveness              of the          1990 promotion               and outreach                  campaign             is
still      being      evaluated,             it     appears          that       the     Bureau             significantly
underestimated               the amount             of public           resistance               to the          census.
Several       evaluations             are underway                 to determine                the         variables            that
contributed           to     the     low mail            response           rate.         The results                 of     these
studies       may shed some light                        on changes             needed         in the           promotion
and outreach           program             or the        census        questionnaire                  itself.


The critical           questions             that        I have just             discussed                 need early            and
sustained          attention.               The dialogue               must begin           now on these
questions          so that         innovative             approaches             for     the      2000 census                   may
be identified              and tested              to determine              whether        they            warrant
implementation,               and if         so,     to what degree.                     We previously
reported       that        the Bureau's              past      census         planning           efforts
generally          started         late;      experienced               delays;          were         incomplete;                and,
perhaps       most importantly,                     failed        to    fully          explore             innovative
approaches.1               Before      discussing              the      importance              of early            planning,
Mr. Chairman,              I want          to emphasize             that        thorough         planning              begins

lTtansition           Series:         Commerce Issues,                      (GAO/OCG-89-IITR,
Nov. 1988).
with     analyzing               the      results         of past      experiences.                     We cannot
overstate            the        importance           of    timely      completion              of     the         evaluations
planned        for         1990 census              operations.

Several        evaluations                 are being         done of         1990 census                operations,
including            the        factors       that        influenced         the mail            response                rate,         the
reasons        why the            Postal       Service         could      not       deliver           millions                of
census      questionnaires,                    and the         Post     Enumeration                 Survey.               We
encourage            the    Bureau          to closely             monitor         the    progress                of    these
evaluations            and complete                  them expeditiously                   so that            the         results
can be used            in determining                     approaches         for        the    future             census.

Now turning            to planning,                  a repetition            of the planning                       process             for
the    1990 census                must be avoided.                   That     is,        we believe                that        after
the    1980 census                was completed,               interest            in planning               for        the        1990
census      waned.               This,      in part,         is understandable                      in view             of how
far    apart     decennial                censuses          are.       We have previously                          reported
that     the Bureau               would      need to complete                 its        plans        for         the     2000
census      early          in     the decade              to have sufficient                   time         for        effective
execution.             In addition,                  we recommended                in    1988 that                the
Secretary            of Commerce ensure                     that     the Census Bureau                       develop               and
implement        an effective                  management            structure,               planning             capacity,
and decision               timetable           in sufficient              time          to help        guide            the
Bureau      in making              decisions              and changes         necessary              for          the     2000

To the Bureau's                      credit,       it     established                  a 21st         century           planning
staff       to begin            the      planning          process             for      the     2000 census.                      This
staff       is    researching                  a range       of options                 for     future            censuses.                 We
believe          that      creating             this      staff          is    an important                 first          step        to
rethinking              the current              census        methodology.                     However,             because             the
reexamination                  of     the      census      methodology                  has broader                 implications
about       the Nation's                 data     collection                  strategies,                 the Bureau              alone
cannot       make these                decisions           without             input          from        a broad          range        of
decisionmakers                 --other          executive           agencies,                 Congress,             data       users,
and statistical                     experts.

One vehicle              for        focusing           discussions              on these              issues         would            be a
commission              or panel            such as the             one proposed                     in    the      Federal
Information              Resources              Management           Act         (S.     1742).             This        act
provides          for      comprehensive                 information                   resources            management                 of
federal          departments                and agencies.                     In February                 1990,      we
testified           in     support             of the      bill's             provision          to create                a
commission              which         would,      among other                  things,          identify             long-term
information              needs and set priorities                                for     information                 in       light         of
changes          in the         economy and society.2                             However,                we expressed
concern          about         the     commission's               broad         mandate          coupled             with        its
short       reporting               deadline.            Such a commission                       could           also         be used
to assess          alternative                  census      methodologies                      and may provide                    some
consistent              high-level              leadership           on important                     census         issues.

2E'ederal Information                          Resources          Management              Act         (S.    1742)            (GAO/T-
IMTEC-90-3,   Feb. 21,                         1990).

    In summary,         Mr.    Chairman,      my main message          today   is   that   we must
    begin    rethinking         the current      census      methodology       now while     the
    experience         of the     1990 census      remains     fresh     and opportunities         for
    change    exist.

    This    concludes         my prepared      statement.       My colleague         and I would
    be pleased         to respond     to questions.