Thread by @lymanstoneky

The Puerto Rico 2020 Census number is.... remarkable. The Census is saying they underestimated Puerto Rico's population by almost 100,000 people!

Their figure also makes an absolute hash of any of the population flows data we have for Puerto Rico.... egads.

By the way, Census' Demographic Analysis program indicated an estimated population of 332.6 million, range of 330.7-335.5. The true value at 331.4 is nearer the low range than the midpoint, so the US population figure should be seen as a slightly pessimistic indicator.

Assume births and deaths were correctly tracked (reasonable assumption) this means that net migration was about 115,000 FEWER people per year than the Census previously believed, between 2010 and 2020.

Back to Puerto Rico, if you just take Puerto Rico's 2010 Census numbers, and add births and passenger arrivals on planes, subtract deaths and passenger departures, April 1 2020 population should have been 3.02 million in Puerto Rico.

So between "what we can document in official statistics" and "who we counted at a point in time" we have a discrepancy of 265,000 people.

For perspective, Puerto Rico only reported 313,000 births total in the decade, and about 313,000 deaths too, so to balance this out on the natural increase side is implausible. It'd imply vital registration of births in Puerto Rico is only like 60%.

And Puerto Rico's life expectancy figures are plausible so it's exceedingly unlikely deaths are being over-reported.

Rather, you have to make the balance up on migration. Air traffic data shows 706,000 net outflows. Census would imply 441,000.

But this is kind of tricky. Because we also have survey estimates: large sample surveys that ask people in Puerto Rico about past migration, and that ask people in the mainland too.

Even if we assume that ZERO Puerto Ricans moved anywhere other than the United States, the survey-based estimates of migration imply a net loss of 543,000.

And folks, that gap is a big miss. The PRCS identified about 275,000 in-migrants between censuses. If there were actually 375,000, that would be kind of.... a big deal. That's not a small miss!

Or if on the other hand it turns out that there were just few inflows into the US from Puerto Rico than the ACS measured.... why was the ACS measurement so wrong? Migration from Puerto Rico is a pretty big estimate! The error bands are not *that* wide!

People are misunderstanding this.

The Census showed a *high* count of Puerto Rico. They're saying that there are 92,000 *more* Puerto Ricans than their prior estimate, and I'm scratching my head about how that can be.

https://twitter.com/ChiNeoliberal/status/1386785825590022151

The error margin on the estimate of inflows into Puerto Rico is about +/- 20,000 across the whole decade, while for outflows to the US mainland it's about +/- 35,000. So the error we're talking about here is nearly twice the size of the mean-lower bound estimate rage.

People told me I should look at New York so I looked at all the states.

Y'all Census massively screwed up somewhere.

My theory of what happened is that seasonal migration patterns were shot to hell by COVID and so they basically measured a non-comparable population to prior years.

Except then why does Hawaii have such a big boost???? I dunno.

But y'all Puerto Rico really is in a league of its own! They had an 8% revision! That's monumental!

But what I'm now even more confused about is.... most states got *upward* revisions vs. the Census Population and Housing Estimates figures.... but the *national* figures came in *below* what Census estimated in their 2020 Demographic Analysis range.

Did the analysts on these two projects just.... not compare notes?

In the 2010 census, the average error vs. last-estimate-vintage was 1%.

For this Census, it was 1.4%.

That is NOT a trivial difference! The Population Estimates evidently declined considerably in quality between Censuses!

(or the Census got worse....)

Absolute average error was greater in the the 2020 round than the 2010 round for 34 states, including many of the largest states.

This despite the fact that we have incontrovertibly gotten *more and better* population data over the years and in theory the science of population statistics has advanced and so should be yielding *better* estimates.

Now look, this Census may be entirely correct and reliable!

Maybe it's the Population Estimates program which is screwy!

But since the estimates program is kind of a huge deal and relied on by federal, state, and local governments for planning and funding purposes we should care if it is deteriorating in quality.

Question maybe @ipums knows the answer to:

Des the Census ever release data on population in a region estimated from Question 1 vs. Question 2 on the form?

Now extremely interested in the Count Question Resolution program!