Hacker News Reader

Comment by aoki | Hacker News Reader

It’s not.

A census literally counts every individual. Census public reports are aggregated for privacy.

This is about spatial disaggregation of aggregated census data. The problem in developing countries (which is what these datasets are used for) is that they often fail to run a complete census (in some cases, for many decades) because it is expensive and/or the government is not functional. So these datasets may not be well-calibrated overall.

The US has no problem running a decennial census, aside from nonresponse by immigrants, conspiracy theory enjoyers, etc.

View on HN

> A census literally counts every individual. Census public reports are aggregated for privacy.

Is there any evidence of how accurately the US census actually does that job? Having spoken with a number of people who were involved with the 2020 census (mostly on a volunteer, local basis) the answer is not "absolutely every person in the country got counted exactly once". There are a number of sources of error that would seem essentially impossible to fully remediate, due to people being complicated and error-prone and the census being largely self-reported:

- People in multiple households being counted twice, e.g. college students being counted both at college and at home.

- People who refuse to participate (one of the census-takers I know had someone wave a gun at her face when she tried to follow up on their household's non-response).

- People who are transient or disconnected enough from the social fabric that they are hard to maintain consistent contact with. I knew someone involved with taking the census for an urban unhoused population and there was no doubt that they both missed and double-counted people; I'm sure it's even more difficult for rural populations that might be even less connected and even more transient.

It's not hard to imagine that, on net and despite there being rules and processes for handling each of those situations, there ends up being regional and/or urban/rural bias in the final counts.

> Having spoken with a number of people who were involved with the 2020 census (mostly on a volunteer, local basis) the answer is not "absolutely every person in the country got counted exactly once".

Do you have any thoughts on how the US census takes your insightful comment's points into account?

I imagine it's got to be quite an involved process given the vast differences in US geography, kinda blows my mind thinking about even basic stuff like age demographics vs. taking into account how many people died on census day.

But then - maybe that's too granular and it all balances out in the end? Or at least, it does if you use special magic sauce the US census has covered?

US rural people tend to trust the government less, so I could definitely see them getting undercounted.

The census should really have an option for refusers that collects an absolute minimum of information. The Constitution only asks that people be counted, not that they be broken down into categories.

At least it's just the short form now. Prior to 2010, one out of six people got the long form, which had ten pages of questions for each person in the household. Here's a pdf: https://www.census.gov/content/dam/Census/programs-surveys/d...

I believe the long form survey you are talking about is the acs survey

If you receive it you are legally required to respond.

https://www.census.gov/programs-surveys/acs/about/top-questi...

Considering cenus data is used to determine representation numbers just counting people could lead inflated numbers if that information can't be verified.

It used to be a "long form" every ten years, now it's the ACS every year.[1] I didn't realize until now that they'd continued the long form in another, um, form.

Technically you're required to respond, with a fine of $100 per question if you don't (with 50 questions). However, in the past half century the government has not prosecuted or fined anyone for failing to respond.[2]

[1] https://www.prb.org/resources/continuity-and-change-in-the-u...

[2] https://www.easidemographics.com/web/do-i-have-to-complete-t...

>US rural people tend to trust the government less, so I could definitely see them getting undercounted

Isn't this more complex? Do they mistrust the census bureau?

I say this because their mistrust seems to ignore politicians who align with their views. The same people who control the government.

For example ask a rural person if they trust "Ron Desantis" or "Greg Abbott"

I'm currently reading The Art of Not Being Governed by James C Scott, which is all about how "hill people" the world over have much in common with each other: like illiteracy, oral traditions, living in difficult geographies, frequent moving, different names for the same geographic features--these "uncivilized" traits deliberately obfuscate themselves and make it more difficult to be found, counted, assessed, taxed, conscripted, and in general forced into submission/subservience to the state they are co-located in. "Rural" doesn't quite cover this group, but certainly has overlap with it.

> US rural people tend to trust the government less

I'm not sure of that at all. Cities have many poeple who are members of black and brown minorities, unpopular religions, undocumented people, and immigrants - those are groups that might prefer to avoid government attention and often are very suspicious.

Trump 45 admin did not prioritize the census and their ramp up wasn’t particularly good as compared to past years. In my city there was like 30% fewer enumerators.

In the US, rural undercounting isn’t really a problem politically, although it has negative impacts on revenue allocation that is population based like sales taxes.

The biggest issue is that poor residents are underreported. This is both an urban and rural issue but from a numbers perspective impacts urban areas more.

> Is there any evidence of how accurately the US census actually does that job?

Yes. The Census uses a post-enumeration survey for this purpose. They found evidence of significant under- or over-counting in 14 states for the 2020 census.

That was considerably worse than the 2010 census, though covid might be partly to blame. Still, afaik, there wasn't anywhere near the rural undercount described in the article.

https://www.census.gov/library/stories/2022/05/2020-census-u...

>A census literally counts every individual.

I'm afraid that this isn't the case; a census is a best-efforts estimate.

There are plenty of people missed for a variety of reason, everything from not wanting to be found, through to simple avoidance. Let alone filling out the forms incorrectly or giving dud answers to the army of amazing people trying to make sense of all the madness.

Edit: realising the above has added to a long list of things that once upon a time I thought were hard-set facts, and nowadays I'm slowly losing my mind over. Coordinates, populated place boundaries, census counts + demographics... I mean, what _is_ an address? Incredibly painful to get to the bottom of that one, at least in the UK where the definition of an address will vary significantly by use case

screams into the void

A census worker wasn't able to get in contact with the people who lived across the street from me. They suddenly moved out one weekend during the first summer of the pandemic and never came back, leaving the house vacant. I didn't know them very well, but the census worker asked me to make my best guess about the ages of each of the adult and child who had lived there.

Did that cause a double count due to them being counted wherever they moved to? Were my guesses correct? Or correct enough? No idea but that's the data the census worker recorded.

The census has an official date to avoid issues like this. Since they lived at the house across the street from you on the official census date of April 1st 2020, the census worker was correct to record them at that address.

They would only have been double counted if they incorrectly reported living at their new address as of April 1st.

There is just no way they count every individual. I live in California and was visiting someone in the central valley. For fun I drove some back dirt roads into some hills and there was a small corrugated sheet metal building style settlement with 20ish people hanging around it a ways back in. It really made me wonder how many people are tucked away. Not like a 24 year old kid who gets a college degree and intentionally lives off grid, but poor/uneducated people who eke out a living somewhere off the beaten path.

Reply on news.ycombinator.com