TIME: Almanac 1990/1990 Time Magazine Compact Almanac, The (1991)(Time).iso/time/cal/pop3

Text File | 1991-05-13 | 15KB | 253 lines

                                                                              ««Two Hundred Years and Counting:

Accounting for Undercount  



   No matter how hard we try, it will prove impossible to obtain an
accurate count of every American as of a single day, April 1, 1990.
No census has ever been 100 percent accurate, and none ever will be.

Table: Apportionment of Membership in House of Representatives, by
       Region, 1790 to 1980

                    Membership (based on census of year)
                                                       Total number
Year      Northeast     Midwest     South     West       of seats

1790           57         (X)         49       (X)          106
1800           76           1         65       (X)          142
1810           97           8         81       (X)          186
1820          105          19         89       (X)          213
1830          112          32         98       (X)          242
1840           94          50         86         2          232
1850           92          59         83         3          237
1860           87          75         76         5          243
1870           95          98         93         7          293
1880           95         117        107        13          332
1890           99         128        112        18          357
1900          108         136        126        21          391
1910          123         143        136        33          435
1930          122         137        133        43          435
1940          120         131        135        49          435
1950          115         129        134        59          437
1960          108         125        133        69          435
1970          104         121        134        76          435
1980           95         113        142        85          435

Note: No reapportionment based on 1920 Census.

Source: U.S. Bureau of the Census, Statistical Abstract of the United
        States, 1988, 108th edition (Washington, D.C.: Government
        Printing Office, 1987), p. 242, which also contains listings
        for each state.

   It may also be too much to expect that every group in our society
will be counted with exactly the same degree of accuracy. Voting
behavior, social attitudes, family structure, income, education, and
many other characteristics differ by age, sex, ethnicity, and race. So
does participation in the national census. It is relatively easy to
enumerate the middle class in their suburbs, because they live at
easily-identified addresses, and are likely to receive and return
their census questionnaire. But it is more difficult to enumerate the
poor and the alienated, because it is both harder to locate them--they
live disproportionately in isolated rural areas and inner-city slums--
and more difficult to enlist their cooperation when they are located.

   Throughout our two centuries of census taking, states and localities
have periodically complained that their areas were undercounted by the
census. From time to time, evidence of serious undercounting has
appeared. For example, the 1870 Census was thought to have seriously
undercounted the population of the South. "Interpolation back from the
1880 figures indicate a potential total undercount of 1.2 million in
the South and a black undercount of .5 million--over 10 percent of the
black population," according to social historian Margo J. Anderson.(21)

   Census Bureau studies after the 1980 Census showed that 99 percent
of whites had been enumerated but only 94 percent of blacks. The
Bureau's studies showed the highest undercount rates were for black
males between the ages of 25 and 54, fully 15 percent of whom were
missed.(22) The Hispanic population was probably also
disproportionately undercounted by the 1980 Census.

   Even before the 1990 Census questionnaire went to the printer, a
lawsuit was filed that seeks to force the Census Bureau to adjust the
actual count using statistical methods, in anticipation of another
undercount that will vary by race. Following the 1980 Census, a total
of over 50 similar suits were filed. In each of these cases, some of
which took nearly the entire decade to complete, the courts found in
favor of the government position not to adjust for undercount.(23)

   Another suit filed more than a year before Census Day, 1990, seeks
exclusion of illegal aliens from the 1990 count. The government's
position is that the Constitution requires that all person be enumer-
ated. Further, even if the courts were to direct the Census Bureau to
exclude illegal aliens, "any proposed method would be likely to have
significant errors that could affect the allocation of one or more
congressional seats."(24)

   While successfully defending its decision not to adjust the 1980
Census tally, the Census Bureau began intensive research into whether
it could develop statistically reliable methods to make adjustments
for undercount by the time of the 1990 Census. If the research con-
cluded that developing and using new statistical adjustment techniques
could achieve a more accurate count of the true number of inhabitants
than not adjusting, then the nation's political leaders and the courts
would have the option of adjusting for undercount.

   Testifying before Congress in July 1986, the Census Bureau's Asso-
ciate Director for Statistical Standards and Methodology at that time,
Barbara A. Bailar, said of the Bureau's 1990 Census efforts: "We will
attempt to take the best census possible and to count everyone, but we
also will do what is necessary to be prepared to adjust the counts if
we determine that adjustment will improve them."(25) Then-Census
Director John G. Keane used the same words in testimony before a Senate
committee in September 1986.(26)

   Dr. Bailar told Congress that the Bureau would have to establish
methods that could accurately measure census coverage for small geo-
graphic levels and a variety of population and housing characteristics.
Further, it would have to establish and publish standards for
evaluating the quality of both the unadjusted and the adjusted census
data. Finally, she said, "We have to implement the adjustment, compare
the adjusted and unadjusted data in light of the standards, and then
release one of the sets of data as the official 1990 Census results."

   If there were two sets of official census results--one before and
one after adjustment--the decision which to use could become hopelessly
politicized. The goal of accuracy would be lost in the calculations of
which localities and politicians would gain or lose by using which of
the two sets of statistics. Any adjustment decision would need to be
firmly agreed to before the outcome of the census were known or else
we risk endless rounds of second-guessing, political maneuvering, and
lawsuits that could seriously weaken the credibility of the census.

   The Census Bureau began its adjustment research in 1984. By the sum-
mer of 1987, the researchers concluded that it might be possible tech-
nically to make statistical adjustments to the census counts that
would come closer to the true number of inhabitants than not making an
adjustment. However, many operational and policy questions remained.

   Because the Bureau appeared to be making progress in developing sat-
isfactory adjustment methods, advocates of adjustment began to take
new heart. "Census Bureau is Urged to Adjust 90 Count to include Those
Missed," headlined a New York Times article on August 20, 1987. It
noted that while statisticians remained divided, Democratic officials
generally supported adjustment and Republicans opposed it. Representa-
tive Mervyn Dymally, a California Democrat who chaired the House
Subcommittee on Census and Population, introduced a bill ordering
adjustment, but it was never brought to a vote.

   The Staff Director of the House Subcommittee on Census and Popula-
tion, TerriAnn Lowenthal, told state and local officials, "Based on
several hearings before our subcommittee and numerous consultations
with the experts, we believe a consensus has emerged that a full-scale
Post Enumeration Survey is a statistically valid and technically fea-
sible method of adjusting the raw census counts. The time has come to
carry out an adjustment of the decennial census figures to eliminate
the known undercounts and overcounts of the population."(27)

   But late in October 1987, the Department of Commerce, the Census
Bureau's parent agency, announced that the government had decided not
to adjust the results of the 1990 Census. "Adjustment may create more
problems than it solves, and may divert resources needed for enumer-
ation," said the Under Secretary for Economic Affairs, Robert Ortner.
(28)

   Additional reasons offered by Ortner included the likely controversy
that adjustment would generate, the suspicion it might create about
the reliability and integrity of census statistics, and the fact that,
even without adjustment, the Census Bureau expected to achieve a 99
percent accuracy rate. Under Secretary Ortner said, "Adjustment is a
threat to the customary process of reapportionment which has been one
of the foundations of our political system. The census count has tra-
tionally been accepted as the best count available regardless of the
political consequences. Adjusting the count may create the appearance
of changing the numbers to achieve a desired political outcome by the
party in office."(29)

   Following this decision, the Census Bureau's Barbara Bailar, who had
guided the research into adjustment methodology, resigned from the
government in protest.

   In July 1988, C. L. Kincannon, Census Bureau Deputy Director, issued
a statement giving the official Census Bureau position on the under-
count issue and the Commerce Department decision. He recalled that the
Census Bureau had been researching the adjustment issue since 1984,
and was continuing its research program in the summer of 1987,
before the Commerce Department made its announcement. Based on this
research, the Census Bureau Director concluded that the Bureau
could "reasonably expect to have the technical capability in 1990 of
assessing the undercount." However, "the decision on whether we would
actually adjust the 1990 Census would occur in 1990."(30)

   Noting that some members of the Bureau's senior staff were not in
complete agreement among themselves on the feasibility of adjustment,
C. L. Kincannon reported that at a meeting of top Bureau officials
held on July 15, 1987, before the Commerce Department decision, a
majority agreed "that an adjustment could not be attempted in time for
legal deadlines without placing the census itself at unacceptable
risk. Most of the group strongly agreed that basic census operations
should not be changed and that the efforts to adjust the 1990 Census
by legal deadlines should not proceed. In light of the above, the De-
partment of Commerce decided not to adjust the results of the 1990
census," Mr. Kincannon said.(31)

   Early in November 1988, when the group of states, cities, and
interest groups sued in Federal District Court in Brooklyn to force
adjustment to the 1990 Census because of undercounting, Barbara Bailar,
now the Executive Director of the American Statistical Association,
filed an affidavit to accompany the suit charging that the Commerce
Department decision was arbitrary, secretive, and "substantively flaw-
ed."(32)


To Adjust or Not to Adjust?

   The two sides of the undercount issue have persuasive proponents.
While both would agree that the nation should obtain as accurate a
count of the population as possible, they disagree about the extent
to which the government should attempt to compensate for the failure
of the census to count the American underclass as well as it counts
the great majority.

   The Census Bureau has been a leader in measuring the degree to which
the census has fallen short of perfection. Not every country's census,
and not every government program, is a forthcoming about its shortcom-
ings. The Bureau has also been a leader in using computerized statis-
tical programs to improve the accuracy of enumeration. The Bureau's
computer programs that impute residents and their characteristics,
when they are known to exist but information cannot be gathered
directly, are statistically sound and proven to be effective.

   The Constitutional requirement for a census was drawn up long be-
fore the era of the computer or the development of demographic methods
or sampling techniques. The Founding Fathers could not have expected
an errorless census. The 1790 Census, which took 18 months to com-
plete, yielded a total of 3,929,326, but the government suspected an
undercount. "I enclose you also a copy of our census, written in black
ink so far as we have actual returns, and supplied by conjecture in
red ink, where we have no returns, but the conjectures are known to be
very near the truth," wrote Thomas Jefferson to George Washington in
1791. "Making very small allowance for omissions (which we know to
have been very great), we are certainly above four millions."(33)

   Error has been part of every enumeration for 200 years and occasion-
ally the errors have been major, but the nation has always agreed to
abide by the findings of the actual enumeration, just as it has agreed
to abide by the outcome of free elections, without statistically
adjusting their results for known demographic differences in voting
behavior.(34)

   There is irony in the fact that the census has reached unprecedented
levels of accuracy, the cries have become louder for even greater per-
fection. Pressures for statistical adjustment of census counts have
grown as more federal money is allocated to states, cities, and local
governments from benefit programs based on census statistics.
Knowing that census statistics contain inaccuracies, legislators
nevertheless placed more responsibility on these statistics.

   The courts will have to decide whether the government is required to
statistically add people to the census count. That decision will be
historic and will influence census taking in the 21st century.


Source:"Two Hundred Years And Counting," The Population Reference
        Bureau, Washington, D.C.