ASA Connect

 View Only
Expand all | Collapse all

Census estimates

  • 1.  Census estimates

    Posted 05-07-2021 13:35

    Anyone have any specific census knowledge of the issue identified in this article? All the changes go one way, so random error doesn't seem like a likely explanation.

    https://www.foxnews.com/politics/house-republicans-biden-census-numbers-estimates-departures



    ------------------------------
    Terry Meyer
    ------------------------------


  • 2.  RE: Census estimates

    Posted 05-10-2021 08:53

    It's really quite simple. The first estimates were rigged by the former administration, which made no secret of doing so. This deliberate malfeasance was caught and corrected by the new administration. The final numbers differ from the early estimates because the early estimates were frauds. 

    Are you by any chance familiar with the term DARVO? Once little known outside the area of sexual abuse - I first learned it in connection with research on the under-reporting of human trafficking - Mr Trump and his affiliates have made it go mainstream. It describes a three-step process: 

    1. Deny: refuse to accept responsibility for the crime - for example, claiming the proposed citizenship question was intended to protect voting rights when the real intent was to cause an undercount 
      What's the big deal about adding a citizenship question to U.S. Census?
      U.S. remove preview
      What's the big deal about adding a citizenship question to U.S. Census?
      U.S. President Donald Trump is making a last-ditch push to add a citizenship question to the 2020 U.S. Census, despite a Supreme Court ruling against it last month and criticism by some states and civil liberties groups that the question is meant to deter immigrants from...
      View this on U.S. >
    2. Attack: in this case, the Fox News article cited above, attacks the integrity of the Census final count when, in fact, it is correct and the earlier estimates deliberately incorrect.  
    3. Reverse Victim Order: the true perpetrator of the abuse claims to be the victim. the former administration's proxies today claim to have lost representation when the truth is they were prevented from stealing over-representation. 
      https://www.oleantimesherald.com/news/nation/house-republicans-question-whether-white-house-interfered-in-2020-census-data/article_4b53d691-3f64-5e52-a171-8b8511d93081.html

      The DARVO process is standard operating procedure for Mr. Trump and his acolytes. They are now attacking the Census and playing the victim because their tampering was caught and the counts corrected. 



    ------------------------------
    David J Corliss, PhD
    Director, Peace-Work www.peace-work.org
    davidjcorliss@peace-work.org
    ------------------------------



  • 3.  RE: Census estimates

    Posted 05-11-2021 08:23
    I think the explanation is that  many Red states put little money into census promotion and the Trump administration did it best to discourage Hispanics and other immigration from responding to the Census.    The results is a likely census undercount in the states ( likely Texas and Arizona)  that did not promote the Census and had relatively large immigrant population. .

    ------------------------------
    William O'Hare
    President
    O'Hare Data and Demographic Services LLC
    ------------------------------



  • 4.  RE: Census estimates

    Posted 05-11-2021 21:33
    William,

    While enjoyable, speculation is outside the realm of our profession. If you re-read my query, I was asking for insight from someone with actual knowledge of the data and methods of the 2020 census.

    ------------------------------
    Terry Meyer
    ------------------------------



  • 5.  RE: Census estimates

    Posted 05-11-2021 09:20
    Having worked in population estimates I can tell you that this is untrue.  The methodology is published and only experiences slight, technical changes over the years.  Every year, the States have the right to review preliminary estimates and to provide evidence-based comments to change them.  The April 1, 2020 estimates were interpolated between the July 1, 2019 and July 1, 2020 estimates.  The goal is to produce the most accurate estimates possible.  Errors mainly arise from the difficulties in estimating migration from all sources and differential decennial census coverage.  The Population Estimates Program works hard to estimate migration and has no control over coverage.

    ------------------------------
    Chuck Coleman
    ------------------------------



  • 6.  RE: Census estimates

    Posted 05-11-2021 16:09
    Dear All,
    The ASA is not aware of any evidence that the previous administration interfered with the US Census Bureau's processing or production of the 2020 population estimates. Similarly, we are not aware of any evidence that the current administration interfered with the production of the 2020 apportionment counts.
    Steve

    ------------------------------
    Steve Pierson
    Director of Science Policy
    American Statistical Association
    ------------------------------



  • 7.  RE: Census estimates

    Posted 05-11-2021 17:19

    David - I'm wondering if you may be confusing the actual 2020 census with the development of 2020 population estimates based on the 2010 census updated with birth/death and immigration records?  The previous administration did try to interfere with the census--by trying to speed up the count even though (because of delays caused by COVID) this would not allow enough time for the Census Bureau to review and edit the data and by ordering the Census Bureau to produce counts of undocumented immigrants. These efforts were challenged in court and were not successful. The current administration explicitly reversed the order to produce undocumented estimates (as well as estimates of citizen voting-age population).  

    Neither the previous nor the current administration interfered with  the production of the population estimates, which are produced on a regular schedule by the Census Bureau's Population Division following long-established procedures. The differences between the population estimates for 2020 and the actual 2020 apportionment counts for states occur in every census.  They may indicate quality problems in either the estimates and/or the census for particular states, which evaluation studies may be able to identify, but there was no "fiddling" of the estimates or the counts.  It is important for everyone to be clear on this point and not to leave any impression that the dedicated staff at the Census Bureau did anything other than their usual highly professional jobs.



    ------------------------------
    Connie Citro
    Senior Scholar
    Committee on National Statistics
    ------------------------------



  • 8.  RE: Census estimates

    Posted 05-11-2021 21:34
    Constance,

    I do not have the obvious deep knowledge you have about the census, but I do take issue with your statement about "interference" with the census by the previous administration. I believe the function of those professionals at the census bureau is to do their professional best to implement/execute the decisions of the executive branch, not to pass judgement on the value of or need for the information. Those at the bureau said that asking a citizenship question would cause an under-count in a subset of the respondents. That's a professional judgment certainly; but, the response should be to devise sampling methods to mitigate the problem, not to characterize the request as "interference". In fact, there had been ample experience with this question in the past and the census bureau's own test showed no under-count would likely occur with such a question.

    To pass judgement on the value or need for specific information is to pass political, rather than statistical, judgement.

    I suppose if some future administration asked the bureau in a decennial census to estimate the number of people keeping rats as pets (to use a facetious and absurd example), I would expect those in the census bureau to do their best to develop questions and methods to accurately and reliably obtain such estimates, rather than to characterize such a request as "interference". Do you disagree?

    ------------------------------
    Terry Meyer
    ------------------------------



  • 9.  RE: Census estimates

    Posted 05-12-2021 07:12
    Many thanks to Constance Citro for the clarification! I was indeed referring to the earlier action where, as she wrote "(t)he previous administration did try to interfere with the census". If this is not the earlier numbers specified in the original post, I stand corrected and apologize for any confusion. I have only referred to action outside the Census Bureau, not their own work which is always excellent and unbiased. I will maintain my long-standing practice of praising the work of the Federal statistical agencies (cf. my article in the March issue of Amstat News) and defending them against outside political pressure. 

    ------------------------------
    David J Corliss, PhD
    Director, Peace-Work www.peace-work.org
    davidjcorliss@peace-work.org
    ------------------------------



  • 10.  RE: Census estimates

    Posted 05-13-2021 07:01
    The Census Bureau, for very good reasons, put out 2020 estimates of the population following their usual methodology, as Connie Citro outlined.  What is new, however, is the fact that they created the 2020 estimates at all.  The release pattern in the last two Censuses was only releasing Census data and the national Demographic Estimates during the period that the Census was being released.

    This year, we have the odd situation of having two sets of numbers one generated by the Census Count and one generated by the estimate system, that is based upon the 2010 Census count as adjusted.  So we have serious Two Number problems\.

    1)  A release of the state population counts used for Congressional Apportionment.  (Released April 26th)

    2) A delayed release of the Census data from the so-called PL94-171 file, which is aimed at redistricting, and includes data down to the blocks, with many iterations of race, taking into account all of the race reporting, including one race, two races, all the way up to six races.  Population and data on the diversity of population both for adults, children and total are derivable from these data.  Besides redistricting these data are used to track population change for any and all geographies down to the block.  Mayor's and citizens should know the population and population distribution for the nation, states, counties, places, villages, minor civil divisions, census designated places, school districts, and their own neighborhood.  Indeed, there are over 50,000 administrative districts of various kinds that the Census Bureau produces data for. (To Be Released Mid-August in one format, and then re-released in September with a new "official" format.)  The impact of the noise injection is still not known, or how that will actually work or whether the Bureau will need to find another approach is being challenged in court.  A recent release of the data from the 2010 Census with noise injected for comparison is not reassuring.  Controversial Census Bureau Plan That Makes Data Less Accurate Goes to Court - Social Explorer
    Social Explorer remove preview
    Controversial Census Bureau Plan That Makes Data Less Accurate Goes to Court - Social Explorer
    As some states celebrate and others reel from the release of the state population figures, a storm continues around the Census Bureau's unprecedented and massively con...Social Explorer provides easy access to demographic information about the United States. We provide thousands of interactive data maps going back to 1790.
    View this on Social Explorer >
     

    3) A panoply of estimates for components of change (at the County Level), population at the state level (from December)  In late may and early June more estimates, including race, age, Hispanic States and sex in five year age bands and others.  None of these rely upon the collected data from the 2020.   See the schedule here:  Schedule

    None of the estimates rely upon the 2020 Census.  So we have a two number problem, and people and statisticians being who they are people are going to wonder which set of numbers, if either, is correct.

    Andy

    ------------------------------
    Andrew Beveridge
    Emeritus Professor of Sociology/ President, Social Explorer/ Independent Redistricting Consultant
    Queens and Grad Center CUNY and Social Explorer, Inc
    ------------------------------



  • 11.  RE: Census estimates

    Posted 05-11-2021 21:33
    David,

    Thank you for your reply, but if you read my query, I was asking for insight from someone who actually had knowledge about the issue.

    ------------------------------
    Terry Meyer
    ------------------------------



  • 12.  RE: Census estimates

    Posted 05-13-2021 12:19
    This feels like a strong deflection away from what I feel is the most important thing I have learned from this thread.

    The DARVO concept explains something I have been puzzling about:  how otherwise quite prominent and very intelligent people are being so easily swayed by the propaganda coming from certain corners of the internet.
     
    What I have noticed is they are getting a continual stream of DARVOs.  Most of the DARVOS are immediately recognized as ridiculous (well at least by folks with the capability to do that), and just dismissed out of hand.  But among the continual stream, some actually hit their mark, and leave folks for whom it is clearly not the case strongly opining that they are the actual victims.

    Thanks David, for pointing out that DARVO is a classical strategy of sexual abusers.  Quite a realization, that (in my opinion) does not reflect well on the current political practitioners.

    Once there is name for it, it is easy to see it in many places.

    We live in interesting times!

    Best,
    Steve


    ------------------------------
    J. S. (Steve) Marron
    University of North Carolina
    Chapel Hill NC
    ------------------------------



  • 13.  RE: Census estimates

    Posted 05-10-2021 09:16
    As usual, the decennial census counts differed from the Census Day population estimates.  Demographers have already started studying them.  Full results will be released by September 30.  (Census Bureau Statement on Redistricting Data Timeline) Since I work for Census, I can't say any more.

    ------------------------------
    Chuck Coleman
    ------------------------------



  • 14.  RE: Census estimates

    Posted 05-11-2021 18:05
    I don't have any specific insight into the census data, but the claim that the discrepancies between the Dec estimates and the recent census counts all are in "one direction" makes sense only if you see the world through quite partisan glasses. As far as I can see, the discrepancies go in two directions: the increases were overestimated (too positive), and the decreases were overestimated (too negative).

    I would not at all be surprised if it turns out the prior administration tried to pervert the Dec estimates. I would be a little surprised if they succeeded. I would be utterly surprised if the current administration both tried and succeeded to pervert the recent census numbers. But as far as I can see, none of those hypotheses are required. Doesn't this look rather like straightforward regression to the mean by noisy Dec estimates? I'm sure there will turn out to be additional technical explanations. But foul play seems, a priori, an unlikely hypothesis (at least to me).

    Here's a relevant article about the NY state results: Why New York State's Population Growth Surprised Experts .

    ------------------------------
    Scott Chasalow
    Director, Head of Translational Bioinformatics Methodology
    Bristol Myers Squibb
    ------------------------------



  • 15.  RE: Census estimates

    Posted 05-28-2021 07:41
    Scott,

    You have stated what I would consider to be a reasonable hypothesis ("regression to the mean").

    So I have a couple of questions: have you performed such a regression? Was the slope estimate 1? What was the R^2? Were there signicant outliers? Which states? How did those measures (fit, slope, outliers) compare to similar analyses in prior years (2010 for example)?

    I would be very interested in your results.

    PS Your statement about others seeing the world through partisan glasses is noteworthy considering you belief ("not at all surprised") that the prior administration tried to pervert estimates, but simultaneously would be "utterly surprised" if the current administration perverted estimates. What's the phrase? Throwing stones while living in a glass house?

    ------------------------------
    Terry Meyer
    ------------------------------



  • 16.  RE: Census estimates

    Posted 05-11-2021 21:34
    Chuck,

    I'm not sure how to interpret your response. Are you saying that studying the difference is a routine process and the results of that study will be released in September? Or are you saying that these results were unlikely enough to warrant a study, the results of which will be released in September? Or are you saying that whatever is released in September will have nothing to do with the unlikely changes from December to February? Or something else?

    ------------------------------
    Terry Meyer
    ------------------------------



  • 17.  RE: Census estimates

    Posted 05-12-2021 08:52
    Private demographers have already started analyzing the numbers, see William O'Hara's comments and numerous media articles.

    The Population Estimates Program always conducts an analysis of the estimates after the decennial release with the goal of improving the estimates.  This will start after all data are released.  This will happen after September's release, because all counts will be released then.  A new complication is the addition of noise to assure differential privacy.  Since I am no longer in the Population Division, I have no more information.

    ------------------------------
    Chuck Coleman
    ------------------------------