Methods for Census 2000 and Statistical Adjustments
David A. Freedman and Kenneth W.Wachter April 2004
Abstract
This article outlines procedures for taking the US census, making adjustments, and evaluating
the results. The census turns out to be remarkably accurate. Statistical adjustment is unlikely to
improve on the census, because adjustment can easily put in more error than it takes out. Indeed,
error rates in the adjustment are comparable to if not larger than errors in the census. The data suggest
a strong geographical pattern to such errors even after controlling for demographic variables, which
contradicts basic premises of adjustment. In fact, the complex demographic controls built into the
adjustment process seem on whole to have been counter-productive.

I have to confess that before the 2000 census, I was skeptical of Mr. Freedman’s and Mr. Wachter’s criticisms of statistical adjustment, which sounded more like carping than fundamental objections to me. However, after the 2000 census was concluded — and it turned out to yield a more complete count than expected due to advertising, outreach and other innovative efforts by the Census Bureau — problems in the undercount adjustment were revealed. I became persuaded that Mr. Freedman and Mr. Wachter were prescient and right to highlight the practical problems inherent in adopting the undercount adjustment.

Together with his colleague Kenneth W. Wachter, Mr. Freedman criticized the Census Bureau’s methods for adjusting the census for the people it missed. They argued: “The census turns out to be remarkably good, despite the generally bad press reviews. Statistical adjustment is unlikely to improve the accuracy, because adjustment can easily put in more error than it takes out.”

I was likewise persuaded and reversed my position, as pug will attest.

There has to be a way to get a more honest count without spying on people or using bogus politically motivated computer models.

you don't need real info

just take the government's word

No long form this time so that should help with response rate. Long form data collected via the ACS which is something of a continuous effort to keep the detailed data the gubment uses for lots of stuff up to date. I plan to respond online again and hopefully it will still accept just number of persons at this location like it did in 2000. Nobody followed up for the rest of the info, probably just filled it in themselves...

I think they'd get a more reliable count if they stuck to basics and quit trying to intrude. If it's so important to get an exact count they shouldn't waste their time trying to get personal info on some percentage of us. They certainly have a lot of fancy statistical tools to play with, but if you can't trust their basic numbers what good is all the rest?

http://factfinder.census.gov/home/saff/mai...ang=en&_ts=

Getting Detailed Data
Decennial Census - taken every 10 years to collect information about the people and housing of the United States
learn more | get data
See the Count Question Resolution Program for information on Census 2000 count corrections.
American Community Survey - an ongoing survey that provides data about your community every year
learn more | get data
Puerto Rico Community Survey - the equivalent of the American Community Survey for Puerto Rico
learn more | get data | en español
Population Estimates Program - population numbers between censuses
learn more | get data
Economic Census - profiles the U.S. economy every 5 years
learn more | get data
Annual Economic Surveys - data from the Annual Survey of Manufactures, County Business Patterns and Nonemployer Statistics
learn more | get data
Special Interest

Counting the peeps correctly enough for me... all the detailed data being self reported can easily be "misreported"

I figure marketing companies do a better job on economic issues anyway. Just count for apportionment and such and let somebody else do the rest. I got stuck with that long form last time and it was a pain in the ass, and they are a bit threatening about filling it out too. I can see why some folks don't reply or feel they have to lie.

I don't want to know how many bathrooms everyone has... or what race they think they are...

I hope the Bureau of the Census counts the 10,000 illegal immigrants here in Chelsea, D-Marx.

It's immaterial what you, individually, want to know. Please, don't be flippant about the rationale behind the questions being asked. You may not agree that the Federal government ought to be involved in the activities and funding of activities that it is currently, but that doesn't mean that GIVEN THAT IT IS it shouldn't have the necessary information in order to effect them properly.

... and having people provide whatever answers they want to the detailed questions provides that necessary information how? There is no reasonable way to check the billions of answers they receive... In the case of race many of the respondents may not even know the "correct" answer let alone want to provide it...

I always insist on the real data... see my signature...

So, your point is that the Census is not a VALID instrument?

No, seriously. There's potential here for you to commit all the "sins" of which you accuse inyer and CR in re 9/11.

you are a joke

I would swear it in court

I'd take the oath

I am saying that asking folks to answer questions which the census bureau personnel are not going to verify (or even be able to verify) and to which the respondent may not even know the "correct" answer is a waste of time...

I am glad to see you agree that Inyer and CR are full of crap regarding 9/11...

Please, stop it. You're liable to be guilty of what YOU ACCUSE inyer and CR in re 9/11.

I am not like you, when it comes to words.

WHy is this such a problem for you?

Let me understand this: So, you would like the Federal government to use data collected by the private sector, to RELY on it in making allocations of taxpayer monies?

You have no idea how much more and more detailed information it could ask for were the criteria not as strict and we all who have an interest in good survey data would metaphorically "kill for". There's no better survey, sampling wise, than the U.S. decennial census.

I am saying that asking folks to answer questions which the census bureau personnel are not going to verify (or even be able to verify) and to which the respondent may not even know the "correct" answer is a waste of time...

WHy is this such a problem for you?

Please, stop it. You're liable to be guilty of what YOU ACCUSE inyer and CR in re 9/11.

I am not like you, when it comes to words.

Fine, fine. Let the federal government make decisions based on NO data, at all.

No, I want them counting citizens and others to apportion Congress. Since they can barely get that right expecting them to figure out how much bathroom access every American has is more than I can expect them to do correctly.



I did the long form. I know what it consists of. I also know how easy it is to lie or do a crummy job.

Folks lie, they make mistakes, they're ignorant. Sometimes they do like me and speed through things just to get it over with.

I figured they needed accurate data so I told the truth.

In the United States since its early history, Native Americans, African-Americans and European-Americans were classified as belonging to different races. For nearly three centuries, the criteria for membership in these groups were similar, comprising a person’s appearance, his fraction of known non-White ancestry, and his social circle.2 But the criteria for membership in these races diverged in the late 19th century. During Reconstruction, increasing numbers of Americans began to consider anyone with "one drop" of "Black blood" to be Black.3 By the early 20th century, this notion of invisible blackness was made statutory in many states and widely adopted nationwide.4 In contrast, Amerindians continue to be defined by a certain percentage of "Indian blood" (called blood quantum) due in large part to American slavery ethics. Finally, for the past century or so, to be White one had to have "pure" White ancestry. (Utterly European-looking Americans of Hispanic or Arab ancestry are exceptions in being seen as White by most Americans despite traces of known African ancestry.) Similar questions were raised about east Asian Americans (i.e. of Chinese American and Japanese American ancestries) could pass as "white" or portrayed as more "assimilated" than other racial minorities, regardless of the fact their racial origins are non-Caucasian or outside of Europe.

Efforts to sort the increasingly mixed population of the United States into discrete categories generated many difficulties (Spickard 1992). By the standards used in past censuses, many millions of children born in the United States have belonged to a different race than have one of their biological parents. Efforts to track mixing between groups led to a proliferation of categories (such as "mulatto" and "octoroon") and "blood quantum" distinctions that became increasingly untethered from self-reported ancestry. A person's racial identity can change over time, and self-ascribed race can differ from assigned race (Kressin et al. 2003). Until the 2000 census, Latinos were required to identify with a single race despite the long history of mixing in Latin America; partly as a result of the confusion generated by the distinction, 32.9% (U.S. census records) of Latino respondents in the 2000 census ignored the specified racial categories and checked "some other race". (Mays et al. 2003 claim a figure of 42%)

I figured they needed accurate data so I told the truth.

Show me how Census employees verify race of respondents...

Show me any proof that the vast majority of respondents understand the race questions, the answers to the race questions, and actually know the "correct" answer to the race questions for themselves and their dependents...

Census is basically voluntary compliance... they assume the answers are correct... even if the folks providing the response don't know the "correct" answer... it turns out even the "correct" number of people residing at this address is a tough question for some...

Census is basically voluntary compliance... they assume the answers are correct... even if the folks providing the response don't know the "correct" answer... it turns out even the "correct" number of people residing at this address is a tough question for some...

It might be news to you that there are no "correct" answers to the race questions.

US CODE TITLE 13 > CHAPTER 7 > SUBCHAPTER II > § 221

§ 221. Refusal or neglect to answer questions; false answers
How Current is This?
(a) Whoever, being over eighteen years of age, refuses or willfully neglects, when requested by the Secretary, or by any other authorized officer or employee of the Department of Commerce or bureau or agency thereof acting under the instructions of the Secretary or authorized officer, to answer, to the best of his knowledge, any of the questions on any schedule submitted to him in connection with any census or survey provided for by subchapters I, II, IV, and V of chapter 5 of this title, applying to himself or to the family to which he belongs or is related, or to the farm or farms of which he or his family is the occupant, shall be fined not more than $100.
( b ) Whoever, when answering questions described in subsection (a) of this section, and under the conditions or circumstances described in such subsection, willfully gives any answer that is false, shall be fined not more than $500.
© Notwithstanding any other provision of this title, no person shall be compelled to disclose information relative to his religious beliefs or to membership in a religious body.

Really? Now you've moved to the number of people living in a particular dwelling, and the size of households.

Have you ever read any Census publications? Do you think you're the only one that's ever thought about these issues?

Yes, No, the folks in charge don't really care about whether the answers are CORRECT as long as they are "correct"...

Yes, No, the folks in charge don't really care about whether the answers are CORRECT as long as they are "correct"...