Capitol Fax.com - Your Illinois News Radar

US Rep. Krishnamoorthi wants Census review of faulty ACS data

Posted in:

Illinois

* Here we go again. Capitol News Illinois last month…

New estimates from the U.S. Census Bureau suggest Illinois is continuing to lose population.

The latest estimates, released Tuesday, pegged the state’s population at 12,671,469 as of July 1, 2021, down by 113,776, or 0.9 percent, from the official 2020 census. […]

The Census Bureau routinely estimates national, state and county populations each year following a decennial census using a variety of data sources. But those estimates have been off in the past.

In 2019, for example, the Census Bureau estimated that Illinois had lost more than 51,000 people since the 2010 census while the official 2020 census showed the state had lost about only 18,000.

This was from the American Community Survey, which has been wrong about Illinois’ population for a decade. The General Assembly used ACS numbers in its first legislative remap, but that was tossed out by the courts for being woefully unconstitutional. The GA had, by that time, already passed a new map using decennial Census data, and that map was upheld.

* US Rep. Raja Krishnamoorthi sits on the Oversight and Reform Committee, which has jurisdiction over the Census Bureau. I chatted with him over the break about the latest ACS numbers and he followed through with a letter to Census Bureau Director Robert Santos this week…

January 13, 2022

Director Santos,

In light of the statistical gaps between the 2019 American Community Survey (ACS) and 2020 Census results, I’m writing to ask that you expand the Census Bureau’s review of the 2020 ACS methodology beyond the impact of the coronavirus pandemic to a broader review of modernizing, updating, and improving the ACS to provide data more reflective of reality on the ground and more closely aligned with the decennial census.

As you know, the ACS is crucial for collecting yearly information on the American public to help local governments make policy decisions to best improve communities across the country and to help businesses adequately serve their customers while planning for the future. I know the U.S. Census Bureau is committed to the integrity and accuracy of both the ACS and the decennial Census and had that belief reaffirmed through the agency’s difficult decision last year to delay the release of 2020 ACS data before releasing it in an “experimental form” and announcing plans for a methodological review to ensure that “the resulting data meet our quality standards.”

In the interests of reaffirming that longstanding commitment to data quality, I am writing to request that you extend your agency’s methodological review of the ACS beyond the COVID-19 pandemic-impacted results to the general practices employed in generating the ACS. As you know, your work is vitally important to our country and while perfect data is impossible, even small errors and margins of error can carry enormous consequences, demanding an aggressive, continuous pursuit of improvement.

The impact of ACS data on public understanding and policymaking is significant, and in recent years, the challenges of inconsistencies between those projections and the Census have sown doubt, confusion, and overreaction. For example, over the course of the last decade, my home state of Illinois’ population decline has been a major story, driving a range of policy debates and disagreements based on ACS projections. However, 2020 Census results suggested that those reports of population decline may have been notably overstated.

Last month, your agency again reported that Illinois’ population was in a state of decline, but the experience of the previous ten years of reports followed by the census has led this result to be met with a degree of skepticism and some diminishment of trust in ACS data. While statistical projections carry the inevitable margins of error and those populations also vary substantially month to month and year to year across the country, your agency’s data is essential to the function of our democracy and economy and so is public trust in that data.

A new methodological review that can address potential shortcomings in general ACS function, analysis, and collection would not only strengthen the quality of ACS data but also highlight areas of need or investment by which Congress can help the Census Bureau meet the data needs of this century. In this interest, I request that the U.S. Census Bureau conduct such a review beyond the scope of the challenges of this pandemic while highlighting avenues through which the agency can improve, and Congress can help it improve, to ensure Americans have access to the highest quality data about our nation.

posted by Rich Miller
Friday, Jan 14, 22 @ 9:37 am

Comments

-But those estimates have been off in the past.-

Yes, that’s why looking at any one single year doesn’t mean much. The only reliable metric is the official count every ten years. The Census Bureau does the best they can do .

Comment by Steve Friday, Jan 14, 22 @ 9:47 am
Long overdue. And when the ACS get estimates wrong year after year the cumulative effect is misperception (deliberately encouraged by many) that the state is losing hundreds and thousands of residents when it is not.

Comment by Roman Friday, Jan 14, 22 @ 9:56 am
I support reviewing the methodology of the ACS, but as it’s a sample it’s always going to have a higher margin of error than the actual Census. Additionally, Illinois’ extra efforts in 2020 to get counted (having a 1% higher self-response rate than in 2010 and 5% better than the national average) probably also contributed to a more accurate picture.

Comment by DuPage Guy Friday, Jan 14, 22 @ 9:59 am
What is the point of publishing data that is so often wildly inaccurate? It stirs up both sides of the political spectrum and any policy based on addressing the issue illuminated by such data is unlikely to be effective. Not to mention it deservedly adds to the overall distrust of government, which is already at record levels. So it is definitely high time to improve the methodology used to generate this data or stop publishing it altogether.

Comment by Captain Obvious Friday, Jan 14, 22 @ 10:00 am
= which has been wrong about Illinois’ population for a decade =

It’s hard to be consistently wrong without an underlying reason. Whether deliberate or not, something is obviously broken. The odds of flipping a coin 100 times and having it come up heads 100 times aren’t impossible, but if such an event happened consistently you would start looking at the coin a bit closer(is it weighted differently, does it have heads on both sides, etc)

This reminds me a bit of a story I saw a few years back on how cheating was detected within sumo wrestling tournaments simply with data analysis. The data that came back initially didn’t make any sense to the researcher, but it kept coming back the same way. There was a pattern of what appeared to be cheating, but it was happening in a way his western eyes couldn’t decipher. It wasn’t until he talked with another sumo wrestler to understand the internal beliefs of sumo wrestling that it became apparent what was happening. I think it was finally determined that over 20 wrestlers were participating in this match-fixing.

Data analysis can uncover things our puny brains might miss, or simply outright dismiss. There’s tremendous value in these collections of data and their analysis, as it also removes the ego we usually put in front of what we are looking at.

So after that long winded rambling, there seems to be something happening here with the census data around estimates. I hope this analysis goes deeper. It may end up being a fascinating cause, at least to the numbers geeks - but it is also important for the function of our government.

Comment by TheInvisibleMan Friday, Jan 14, 22 @ 10:01 am
I’m a Republican but I respect the way that Raja gets down in the weeds on policy issues and actually takes his job seriously unlike many other current members of the House. We need more thinkers like him in both parties.

Comment by Middle Way Friday, Jan 14, 22 @ 10:07 am
There used to be 1, 3, and 5 year ACS estimate, but the 3 year ones were axed in budget cuts.

Accurate ACS numbers are important because they are used for a variety of federal funding decisions amounting to hundreds of billions each year.

Comment by Google Is Your Friend Friday, Jan 14, 22 @ 10:11 am
The US Census numbers are like buying a GIA graded diamond from the local jewelry store that has been in business for generations. The American Community Survey numbers are like getting a diamond from someone off Craigslist and trading cash for it behind the 7-11. Are you going to get an actual diamond instead of glass? It’s possible, but not bloody likely.

Comment by Da big bad wolf Friday, Jan 14, 22 @ 10:11 am
There are a lot of issues with how the Census counts people that should be looked into while he is at it. For instance, how prisoners are counted where the prison is located when they cannot vote and cannot access services in the community.

Comment by City Guy Friday, Jan 14, 22 @ 10:18 am
Last year I think there was discussion on this board about how there is a new type of metro area developing with the unusual characteristic of losing population, but increasing in productivity/GDP.

Last week I was reading some research out of Ohio that seems to be showing there are also 5 new types of rural areas developing, each with their own unique population trends. To avoid cluttering the site with urls, just search ‘There is no longer one rural america’ to read the details.

I have to wonder if the census estimate formulas are adapting quickly enough to the way our country is changing.

At the very least, these misses seem to be pointing to an underlying and significant change in the way the US population moves around.

Comment by TheInvisibleMan Friday, Jan 14, 22 @ 10:26 am
===Last month, your agency again reported that Illinois’ population was in a state of decline, but the experience of the previous ten years of reports followed by the census has led this result to be met with a degree of skepticism and some diminishment of trust in ACS data. While statistical projections carry the inevitable margins of error and those populations also vary substantially month to month and year to year across the country, your agency’s data is essential to the function of our democracy and economy and so is public trust in that data.===

I grabbed so much is because it’s not a “nuance” or “alternative fact” choice.

The reason phonies continue to push such bad data is because nuance and “fact choice” only needs the fuel of inaccurate data.

It’s far bigger than “do better”, it’s bigger to the idea of honesty.

Comment by Oswego Willy Friday, Jan 14, 22 @ 10:34 am
With crowd sourced data to supplement field surveys, it should be easier than ever to make more accurate projections than, say, 10 years ago. It’s just a matter of avoiding GIGO and using the right methodologies and qualified/trained people.

Comment by Six Degrees of Separation Friday, Jan 14, 22 @ 12:17 pm
The Capitol News Illinois article from last month referenced above seems to be reporting on PEP data, not ACS data.

Comment by Anonymous Friday, Jan 14, 22 @ 1:08 pm
I tend to agree with 10:07 as the Congressman always seems dialed-in.

Comment by City remap Friday, Jan 14, 22 @ 3:33 pm