The US Census Bureau thread

[edale]

How do people know the Census estimates/counts are wrong? What other data do we have to definitively state that they undercounted? I know the Census Bureau will release corrections, but it seems people are confident that they get it wrong more than these corrections would indicate.

[Steely Dan]

Originally Posted by edale

How do people know the Census estimates/counts are wrong?

History.

The 1990s:
CB estimated Chicago lost ~100K
When they counted in 2000 the city grew by ~100K
The miss: 200K


The 2000s:
CB estimated Chicago gained ~100K
When they counted in 2010 the city shrank by ~200K
The miss: 300K


The 2010s:
CB estimated Chicago was basically stagnant
When they counted in 2020 the city grew by ~50K
The miss: 50K

[Crawford]

Originally Posted by edale

How do people know the Census estimates/counts are wrong?

At least in the older, more urban places, their annual estimates and the official decennial counts aren't close. So they can't both be right. And I'm not talking about small rounding errors. They were once off like 10% in NJ.

And even if the annual estimates were right, the takeaways are wrong, since the annual estimates start from the decennial (enumerated count) base.

[edale]

Ah, ok, so the issue is with the estimates, not the actual counts. That makes sense.

[Yuri]

Originally Posted by Steely Dan

HARD disagree.

That's because you have insanely high standards.

All the other offices operate like this: city X grew by 3% on the past ten years and the country X 6%. If on the next the country is projected to grow 4%, the city will be on 2%. It's just a straight line on a chart. It's that sophisticated.

[Steely Dan]

Originally Posted by Yuri

That's because you have insanely high standards.

Not really.

I can appreciate how fucking difficult it is to estimate highly complex systems that literally have millions of moving pieces.


It's just that the CB's terrible track record has brought me to the belief that their population estimates for the big, urban, messy cities are essentially worthless.

I'm not arguing that anyone else could necessarily do it better.


Only 8 more years to go until the census 2030 city counts are released!!!

[iheartthed]

Maybe the USCB should let cities submit their own tallies during the non-census years, similar to what the FBI does for annual crime stats. If a city opts out of submitting their own stats then the USCB could just give them a shitty estimate.

[Crawford]

Originally Posted by Yuri

That's because you have insanely high standards.

What? If NJ has (say) 9.5 million, and the Census comes up with something almost 1 million off, you wouldn't say that's slightly problematic?

Why even have a Census bureau then? Why even have annual estimates? Barring some cataclysmic event, any idiot can guess that a place with officially X population will generally have something roughly around X population 1 year afterwards.

[edale]

Originally Posted by Steely Dan

Not really.

I can appreciate how fucking difficult it is to estimate systems that literally have millions of moving pieces.


It's just that the CB's terrible track record has brought me to the belief that their population estimates for the big, urban, messy cities are essentially worthless.

I'm not arguing that anyone else could necessarily do it better.


Only 8 more years to go until the census 2030 city counts are released!!!

I guess this begs the question why you think the decennial counts are accurate given the distrust you have for the estimates? There are potential issues with the methodology used to conduct those counts, too, no?

[Steely Dan]

Originally Posted by edale

I guess this begs the question why you think the decennial counts are accurate given the distrust you have for the estimates?

Neither of them are likely accurate.

But I place WAY more faith in the actual counts given the hurculean effort that goes into obtaining that data.

[Crawford]

Originally Posted by edale

I guess this begs the question why you think the decennial counts are accurate given the distrust you have for the estimates? There are potential issues with the methodology used to conduct those counts, too, no?

True, but you would generally trust an enumerated count over a sampled estimate. The enumerated count is the official number.

And the sampled estimate is based on the enumerated count so even if the sampled estimate were correct, the conclusions are off. City A has X population but if (say) you overestimated the official count, the unofficial estimates would show population loss even if it were inaccurate (bc your initial base count was too high).

[iheartthed]

Originally Posted by edale

I guess this begs the question why you think the decennial counts are accurate given the distrust you have for the estimates? There are potential issues with the methodology used to conduct those counts, too, no?

Far more goes into the decennial census, including sending forms to every residence, and sending out actual workers to validate data.

[Yuri]

Originally Posted by Steely Dan

Not really.

I can appreciate how fucking difficult it is to estimate highly complex systems that literally have millions of moving pieces.


It's just that the CB's terrible track record has brought me to the belief that their population estimates for the big, urban, messy cities are essentially worthless.

I'm not arguing that anyone else could necessarily do it better.


Only 8 more years to go until the census 2030 city counts are released!!!

What I'm saying is they look to stuff occuring on real time, like new housing permits, new electric connections, births, deaths or things you know better than me. That's why they can state that Chicago lost people in 2023 and gained in 2024. An incredibly detailed work.

And I can't stress enough: Americans move like crazy, making such accurate estimates a real achievement.

Originally Posted by Crawford

What? If NJ has (say) 9.5 million, and the Census comes up with something almost 1 million off, you wouldn't say that's slightly problematic?

Why even have a Census bureau then? Why even have annual estimates? Barring some cataclysmic event, any idiot can guess that a place with officially X population will generally have something roughly around X population 1 year afterwards.

That's a very specific case and we can jump to conclusions based on it. What about tons of cases they're spot on?

[Steely Dan]

Originally Posted by Yuri

That's why they can state that Chicago lost people in 2023 and gained in 2024.

You're too much.

[Yuri]

Originally Posted by Steely Dan

You're too much.

If you cannot appreciate they do something very different from simply projecting into 2020-2030 exactly what happened between 2010-2020 (as everybody does), well, that's up to you.

The fact remains no stats office in the world put so much efforts on doing real estimates even to the city level, taking into account many variables that no one even bothers to look at.

They are obviously not omniscient, but they're definitely not "shit" as you described.

[Steely Dan]

^ if you want to believe that USCB population estimates have meaning, that's awesome.

But it's a leap of faith I will 1,000,000% NOT be following you on.

At this point, all we can do is agree to disagree.

[ChiSoxRox]

Edit: Don't take the bait.

[Yuri]

Originally Posted by Steely Dan

^ if you want to believe that USCB population estimates have meaning, that's awesome.

But it's a leap of faith I will 1,000,000% NOT be following you on.

At this point, all we can do is agree to disagree.

US Census Bureau captured the loss of inhabitants in many states and metro areas during the pandemic years despite all those areas having post growth between 2010-2020, some of them with very robust growth (San Francisco, Boston, New York).

And I guess everybody agrees those places did in fact lose population between 2020-2022 and US Census Bureau showed us exactly that. No other office would be able to do it. And they deal with countries where domestic migration is much, much lower.

They're not God, obviously, but they can show us trends as good as our own urban nerds guesstimates, something we won't find anywhere.

[3rd&Brown]

Originally Posted by Yuri

If you cannot appreciate they do something very different from simply projecting into 2020-2030 exactly what happened between 2010-2020 (as everybody does), well, that's up to you.

The fact remains no stats office in the world put so much efforts on doing real estimates even to the city level, taking into account many variables that no one even bothers to look at.

They are obviously not omniscient, but they're definitely not "shit" as you described.

It's problematic that its so off because our federal government allocates resources to state and local governments based on these estimates. So basically, the big cities get the right amount of allocation in year 1 of the census period then from there they get short changed. Conversely, the "fast growth" cities are overallocated.

It also feeds into an ongoing narrative about America's major (generally Liberal) urban centers...that they're all in decline etc. Year in and year out conservative rags like Forbes Magazine and the Wall Street Journal use the estimates to write fluff pieces about how the population changes are evidence of the efficacy of conservative economic policies and the failure of liberal ones...and those are the sound bites that take hold in the American political and economic psyche...then like every ten years, we say, oh kidding...it wasn't real. Then just revert back to the narrative the next year when the estimate shows another decline.

So if 9 out of ten years the estimates are wrong and 1 out of ten years the count rectifies the issue, the average person doesn't understand the process and the history so they just think the 1 year growth is the exception and the city is generally in decline. When in reality, it was never in decline and the 1 year reset is often so massive it offsets a decade of erroneous counts.

[Yuri]

Originally Posted by 3rd&Brown

It's problematic that its so off because our federal government allocates resources to state and local governments based on these estimates. So basically, the big cities get the right amount of allocation in year 1 of the census period then from there they get short changed. Conversely, the "fast growth" cities are overallocated.

It also feeds into an ongoing narrative about America's major (generally Liberal) urban centers...that they're all in decline etc. Year in and year out conservative rags like Forbes Magazine and the Wall Street Journal use the estimates to write fluff pieces about how the population changes are evidence of the efficacy of conservative economic policies and the failure of liberal ones...and those are the sound bites that take hold in the American political and economic psyche...then like every ten years, we say, oh kidding...it wasn't real. Then just revert back to the narrative the next year when the estimate shows another decline.

So if 9 out of ten years the estimates are wrong and 1 out of ten years the count rectifies the issue, the average person doesn't understand the process and the history so they just think the 1 year growth is the exception and the city is generally in decline. When in reality, it was never in decline and the 1 year reset is often so massive it offsets a decade of erroneous counts.

Brazilian tax system is even more centered on federal government and they do the money sharing based on the yearly estimates. By law, Brazilian Statistical Office must delivery their yearly estimates to the Ministry of Planning.

In a sense, our estimates down here are more "politically correct" on the sense they simply repeat what happened on the previous decade, with some reducers based on the country demographics (which is much easier to predict than the one from a city). They play safe, they're "neutral".

However, that shows cities under endless migration boom when it clearly it had ended or the opposite, some cities are showed dragging while they're already on a better shape. I myself identified many of those cases during the 2010's and when the 2022 Census came, I was right.

And things are not so tragic here as domestic migration rate is much lower than in the US, coupled with only a small surplus of international migration.

US Census Bureau at least try to address changes on trends inside a given decade.