Subject: URGENT! PUMS 2000
From: Caroline M. Hoxby (choxby@kuznets.fas.harvard.edu)
Date: Sun Jun 04 2000 - 05:11:01 EDT
Dear Dan,
For those who are interested (LS and PE email lists probably), I am
forwarding more information about the changes to the 2000 PUMS.
Apparently, user emails/letters will be helpful in preventing problematic
changes to the PUMS. Could you forward this as you see fit? Caroline
Hoxby
---------- Forwarded message ----------
Date: Thu, 25 May 2000 10:28:26 -0500 (CDT)
From: Census 2000 <census2000@atlas.socsci.umn.edu>
To: survey-list@hist.umn.edu
Subject: URGENT! PUMS 2000
Dear PUMS Survey respondent:
We reported the results of the survey to the Census Bureau at a May 22
meeting in Washington, and I am optimistic that we made real headway in
persuading Bureau staff that detailed subject categories are essential to
users. The plan to suppress detail, however, is in some ways more extreme
than we thought at the time we wrote the survey. For example, no specific
countries of birth would be provided under the Bureau's proposal.
Instead, we would only be able to identify continents of birth.
The full report, including the details of the current Census Bureau plan,
is posted at http://www.ipums.org/~census2000. As described in the
report, we have proposed that the Bureau consider a more moderate approach
to reduce the risk of disclosure.
We believe that the proposed changes would severely compromise the social
science and public policy infrastructure of the United States without
materially reducing the risk to respondent confidentiality. No final
decisions have been made, however, and the Bureau is actively seeking
input from users. I feel confident that if we make our concerns known,
the Bureau will moderate these cuts. Therefore I urge you to write to
Kenneth Prewitt and Paula Schneider, the key executive staff at Census
(addresses below). Your letter could go into detail about the kind of
research these changes would preclude, but even a brief letter or email
expressing your concern would be extremely valuable. We will send a
powerful message about the value of the PUMS to user community if everyone
who responded to the survey also sends a letter.
Kenneth Prewitt, Director Bureau of the Census
Federal Office Bldg. 3, Room 2049
4700 Silver Hill Road
Suitland, Maryland 20233
kenneth.prewitt@ccmail.census.gov
Paula J. Schneider Principal Associate Director for Programs
Bureau of the Census
Federal Office Bldg. 3, Room 2037
4700 Silver Hill Road
Suitland, Maryland 20233
paula.a.schneider@ccmail.census.gov
We would also appreciate it if you would send us a copy of your letter.
Task Force on the 2000 PUMS
Minnesota Population Center
537 Heller Hall
271 19th Avenue South
Minneapolis, MN 55455
census2000@hist.umn.edu
FAX: (612) 624-7096
I know this is a busy time of year for many of you, but speed is of the
essence: key decisions may be made at any time. If we do not respond, we
will have no right to complain later.
Thank you for your help, and please feel free to forward this message to
any individual or list you think might be interested.
Yours,
Steven Ruggles
ruggles@hist.umn.edu
Chair, ICPSR Census 2000 Advisory Committee
Principal Investigator, IPUMS Project
------more follows from Steve Ruggles-----
Dear PUMS Users:
We have now had a chance to do some analysis of the reduced-detail subject
classifications proposed by the Bureau, and they just don't seem to make
much sense. The Census Bureau has a long tradition of ground-breaking
research into methods for confidentiality protection. The plan proposed by
the Bureau at the May 22 conference, however, shows no sign of such
innovative thinking. On the contrary, the proposal to reduce detail in the
PUMS appears to be to be hastily conceived with no systematic underlying
principles.
According to the presentation at the May 22 conference, the goal was to
collapse categories with fewer than 0.5% of the population, which works out
to 1.2 million persons in 1990. The plan as presented, however, calls for
collapsing many much larger categories and yet leaves some very small
categories intact. For example, consider the case of occupation. The
Bureau is proposing to collapse occupational groups like teachers,
construction trades, and motor vehicle operators into single huge categories
containing upwards of five million people each, while retaining distinctions
of archivists and librarians (27,000 persons in 1990 nationally) and social
scientists (43,000 nationally). Many large occupational groups like child
care workers (1.3 million persons) will no longer be identifiable in the
PUMS if the changes go through.
In some cases, extremely small groups will be identified under the proposed
plan. For example, the number of Native American tribes identified would go
from 25 to 39, and some tribes representing fewer than 6,000 persons in 1990
will be identified. The proposed birthplace codes illustrate the basic
inconsistencies. Foreign countries of birth are not identified under the
Census Bureau plan, but U.S. Possessions are. Thus the Mexican-born,
numbering 4.5 million in 1990, are not identified, but American Samoans,
with 12,000, are identified. And inexplicably, persons born in the Northern
Marianas, never before identified, will now be given their own code, putting
them on the same footing as those born in Europe, Asia, and Latin America.
Despite repeated inquires, no one at the Bureau has been able to offer any
rationale for these kinds of anomalies, although one staff member
acknowledged that Native American tribes have to be identified for political
reasons. It is increasingly clear that there is no scientific basis for the
proposed classifications.
It appears that the Census Bureau fears about respondent disclosure in the
PUMS are based largely on experiments carried out by the Census Bureau
Disclosure Review Board that involve linking very small cells in the
aggregate census tables (STF 1 and 3) to the PUMS. In theory this could
allow one to identify the census tract of an individual case in the PUMS and
therefore poses a confidentiality risk. In reality, however, this linking
procedure would not allow a positive identification of tract, since both the
aggregate files and the PUMS were subjected to confidentiality edits to
protect persons in small cells. Apparently some at the Bureau do not think
that the existing confidentiality edits are sufficient. The irony is that
the reductions in subject area detail proposed by the Bureau will not
ameliorate the risk posed by linking in the slightest, since even under the
Bureau's plan the PUMS will still include as much subject detail as do the
summary files.
The Bureau has legitimate concerns about the Congressional criticism of the
long form and about the reduced mail-in response for the long form
questionnaire. But haphazard and excessive reductions in the detail
available to scholars and planners will do little or nothing to protect
respondent confidentiality. Nor is it plausible that such technical changes
would have any effect on public or Congressional acceptance of the long
form.
It looks like there may be some softening of the Bureau's position on these
issues, and the critical decisions may come in the next few weeks. If you
agree with me on the need to preserve as much detail as possible and you
haven't done so already, I urge you make your views known to the Bureau.
The relevant addresses are on our web site,
http://www.ipums.org/~census2000/. If you can think of anyone else who
might be interested in the issue and might not have heard about it yet,
please let them know.
Thanks,
Steve
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Steven Ruggles
Minnesota Population Center
537 Heller, University of Minnesota
271 19th Avenue South
Minneapolis, MN 55455
FAX: (612) 624-7096
PHONE: (612) 624-5818
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
