My Two Census

Formerly the non-partisan watchdog of the 2010 US Census, and currently an opinion blog that covers all things political, media, foreign policy, globalization, and culture…but sometimes returning to its census/demographics roots.

Views from the Census Bureau: A critique of coverage

Here is a contribution from a Census employee. MyTwoCensus grants anonymity to Census Bureau employees who wish to tell their stories or express their views on issues surrounding the 2010 Census without fear of repercussions. Census employees who would like to contribute to the site can find details on our contact page. The following does not reflect the opinions of MyTwoCensus or its editors.

Here is my complaint about a very poor piece of journalism published by The Economist in their special edition titled “The World in 2010″. Perhaps the story about the 2010 Census, headlined “Counting Heads” was meant to be nothing more than infotainment. If so, for those of who who take the census seriously, this story by John Grimond should have carried a warning against reading it.

The writer’s failure to discuss the graphic on the same page as the article tells us that he probably did not understand it. The graphic, a map indicating which states are like to gain or lose representation in Congress based on expected changes in the size of their populations, is sourced to the Population Reference Bureau. Is the Population Reference Bureau’s web site a better source of information about the 2010 Census than The Economist?

The reporter’s ignorance of statistics is another problem. He writes “The longer form used to be sent only to a selection of households, from which general conclusions were inferred.” More precisely, the long form used to be sent to a sample of households, from which estimates were made. Grimond also writes “The details…are now gathered by an annual survey of a small proportion of the population”. Once more, Grimond should have used the word sample instead of the word proportion.

Grimond says “…gerrymandering is now elaborated by computers, not pens…”. Decades ago, this was news. Anyway, computers do not gerrymander redistricting; people use computers to gerrymander redistricting. As even Grimond reminds us, pens can can be used to gerrymander redistricting.

In Grimond’s last paragraph, he tells us that the 2010 Census “will end with the news that the resident population stands at  311,349,543 at year-end, give or take a few hundred thousand.” While Grimond appears to have the April 1 population count confused with  the year-end delivery of that count, that is not the bigger problem. Given the insufficient work done to prepare for the 2010 Census, it is irresponsible to state a population projection without giving its source and explaining how it was arrived at. Both the 2006 Census Test and the 2008 Census Dress Rehearsal were incomplete tests, reduced in scope by budget cuts. Census Bureau Director Robert Groves has told us of a systems test. There is no reason to think this systems test involved any real data. If it was an actual test, you would have been given some information about the test.

The 2010 Census will deliver a massive amount of messy data. Duplicate enumerations, unforeseen record linkage results, missing data and mistakes, perhaps badly covered up, mean that determining the population count may not follow a plan. Let no one suggest a population count to the Census Bureau now.

Grimond’s qualification, “give or take a few hundred thousand”, thoughtlessly suggests that the 2010 Census error will be smaller than Census 2000 error. Census error cannot be known in advance.

If you missed Grimond’s story on the 2010 Census, don’t bother reading it.

Tags: , , ,

One Response to “Views from the Census Bureau: A critique of coverage”

  1. Anonymous CL Says:

    Good post, except I’m not sure why it went on this tangent: “Census Bureau Director Robert Groves has told us of a systems test. There is no reason to think this systems test involved any real data. If it was an actual test, you would have been given some information about the test.”

    I can tell you about that test. It was a load test of computer software that will be used in the LCOs and RCCs for various parts of the 2010 Census work – personnel management, payroll, assigning work areas to field ops teams, etc (the population statistics will only be tabulated in the National Processing Center and HQ, so there’s no software doing that at this level). This load test was performed twice in December. The software programs are being developed on a very last-minute schedule, so they needed to be tested to find where the bugs are, and show that the servers running them need more resources to handle the demand on them. The test was done by having people in all of the LCOs and RCCs simultaneously try to use the software programs and enter made-up example data into the programs, to simulate the load (demand) there will be on the systems when all the offices are running at peak busy-ness later this year.