To Give You a Size of the Immense and Growing Size of Illegal Immigration

[jonatha]

If the census estimates that a state has 2,900,000 million people, and it undercounted black and hispanic people by 200,000... what is the effect on the number of seats that state has in congress?

Zero.

You need 700,000 people to qualify for a representative...

 

[Emily Lake]

If the census estimates that a state has 2,900,000 million people, and it undercounted black and hispanic people by 200,000... what is the effect on the number of seats that state has in congress?

Zero.

You need 700,000 people to qualify for a representative...

I was under the impression it was 750,000.

But hey, okay. Let's revise that:

If the census estimates that a state has 2,700,000 million people, and it undercounted black and hispanic people by 200,000... what is the effect on the number of seats that state has in congress?

 

[jonatha]

If the census estimates that a state has 2,900,000 million people, and it undercounted black and hispanic people by 200,000... what is the effect on the number of seats that state has in congress?

Zero.

You need 700,000 people to qualify for a representative...

I was under the impression it was 750,000.

But hey, okay. Let's revise that:

If the census estimates that a state has 2,700,000 million people, and it undercounted black and hispanic people by 200,000... what is the effect on the number of seats that state has in congress?

Still zero?

 

[Don2 (Don1 Revised)]

If the census estimates that a state has 2,900,000 million people, and it undercounted black and hispanic people by 200,000... what is the effect on the number of seats that state has in congress?

Zero.

You need 700,000 people to qualify for a representative...

I was under the impression it was 750,000.

But hey, okay. Let's revise that:

If the census estimates that a state has 2,700,000 million people, and it undercounted black and hispanic people by 200,000... what is the effect on the number of seats that state has in congress?

Still zero?

She's trying to say the true count is 2.9M even though, the census does a count and the undercount is an estimate....and she's also assuming that after reapportioning, redistricting, and voting, that's a missing Republican congress critter that'd be there.

 

[jonatha]

I certainly would like Emily Lake to reconcile her acknowledgement that a state needs something on the order of 700,000 residents to change the number of Representatives it has and her list of estimated counting errors showing only one state (Florida) with a deficit approaching that number and one state (New York) with a similar surplus with her continued insistence that there are 6 more blue Congressmen than would be the case if all the counts were exact...

 

[Don2 (Don1 Revised)]

If the census estimates that a state has 2,900,000 million people, and it undercounted black and hispanic people by 200,000... what is the effect on the number of seats that state has in congress?

Zero.

You need 700,000 people to qualify for a representative...

I was under the impression it was 750,000.

But hey, okay. Let's revise that:

If the census estimates that a state has 2,700,000 million people, and it undercounted black and hispanic people by 200,000... what is the effect on the number of seats that state has in congress?

That's already a revision on the table based on Arkansas.

 

[Don2 (Don1 Revised)]

But here's where the census bureau talks about the over/under counting, and the percentage error:

2020 Census Undercounts in Six States, Overcounts in Eight

The 2020 Post-Enumeration Survey provides an estimate of the undercount or overcount results for each state and by census operation.

www.census.gov

According to the PES, which states had undercounts?​

• Arkansas (-5.04),
• Florida (-3.48),
• Illinois (-1.97),
• Mississippi (-4.11),
• Tennessee (-4.78), and
• Texas (-1.92).

And overcounts?​

• Delaware (+5.45),
• Hawaii (+6.79),
• Massachusetts (+2.24),
• Minnesota (+3.84),
• New York (+3.44),
• Ohio (+1.49),
• Rhode Island (+5.05), and
• Utah (+2.59)

14 out of 50 states had statistically significant errors. 5 out of 6 of the undercounted states are traditionally red; 7 out of 8 of the overcounted states are traditionally blue.

I still want to know what their measure of statistical significance is. This was based on polls vs census data--they identified the number of households polled, implying that they were random across the country. If so they simply don't have enough samples in the smaller states to be declaring these significant without setting a very low bar for significant. And if they set a low bar then this becomes simply expected noise.

Okay, so, first, if you look at the magnitudes of the "undercount" or "overcount" estimates vs the populations of the states, you DO see a correlation so that smaller states have larger "error." It has R^2 of .34 which isn't strong but isn't near 0 either. So it's conceivable there is an effect there. I was also searching for the significance and didn't find the threshold they are using...just lots of mentions of statistically significant things...

 

[Shadowy Man]

If there is an accurate enough account to show the census undercounted why don’t we use *that* count instead of the census? Am I missing something here? Sorry, I haven’t followed all the posts in this thread

 

[Loren Pechtel]

On the other hand... House seats aren't apportioned on the basis of melanin content or ancestry. They're apportioned to the STATES. I did not suggest any kind of partisan conspiracy in the Census count. I do, however, OBSERVE that the miscounts result in a different portion of house seats being granted to those states where the count was off by a statistically significant amount.

There still has been nothing to address whether it's the census or the survey that's at fault.

You mean OTHER THAN THE CENSUS BUREAU saying that the error is in the census?

You still have failed to address my point: Why should this poll be considered more accurate than the actual census?? And given the limited information we have on sample sizes it looks like they must be using a low bar for p to count those as significant.

We have had supposedly non-partisan officials engage in some very biased acts, given what we can discern from that page it certainly looks like this is another such.

 

[Loren Pechtel]

So let's try this rewording:

A census does a full count of a state's population as 3,011,524. (See how this is better? The census isn't rounding to the nearest hundred thousand or estimating.) According to a survey which might be wrong and have systemic issues, an estimate of an undercount for that state is 151,781 which is a midpoint of a confidence interval.

Do you disagree so far with the changes in wording?

And what is that confidence interval, and what p value is used in calculating it? While reporters will almost certainly omit such things any proper poll will have this information. Why was it not listed???

 

[Don2 (Don1 Revised)]

...
I was under the impression it was 750,000.
...

You might think that you take the total population and divide it by 435, the set number of reps since 1913. This would give about 330M / 435 ~= 761K. However, it doesn't work that way. It might even be close most of the time. One might go ahead and imagine apportionment methods, but it's actually more complicated than intuitive arithmetic and the process has a history that is centuries long.

Problems began to appear early with intuitive methods. Like rounding off the numbers to the nearest whole number. Divide 19 reps among 3 states each having 6.33 delegates. Do you round up, round down/truncate? If you round up, they are all 6, not 19, but there are more complicated scenarios that you can google which gave rise to very insane stuff.

First, you need to meet a Constitutional mandate that each state has to have at least 1 rep. So, all these reps are allocated in the complicated apportionment process. That leaves 385 reps to be distributed among the 50 states in some way based on populations. They do this 1 rep allocation at a time. They assign each state a priority which is the ratio of the state population (P) to the geometric mean of its assigned reps so far in the process and the number it would get if it got this one (n+1). So priority = P / sqt(n * (n+1)). In the 51st round, California will take that spot because all denominators are the same for all states and California has the highest P. Texas might be the next one since California's denominator has increased dramatically in comparison to its last one. The process continues until all 435 spots are assigned.

This method is called the Huntington-Hill method and was alleged to be unbiased when it was implemented in the 40s. However, analysis since then has shown it is biased toward smaller states.

That all said, I'm not going to go through that process to try to prove anything. That would be a bit of work to do.

Suppose we tried to assess this imperfectly by examining the intuitive ways to do apportionment. The 2020 US Census population was 331,464,948. Divided among 435 delegates, we'd get an average of 761,988. I've seen different figures near this floating around the Internet so let's hope we do not observe numbers on the edge.

Suppose we did look at Arkansas. Its 2020 census population was 3,011,524. The midpoint of the undercount confidence interval is 151,781. So add the numbers together and you get 3,163,305. Now, divide each by 761,988. You go from about 3.95 delegates to about 4.15 delegates. Both of these easily round to 4. And if you check the apportionments for Arkansas, you will find that they do have 4 currently. There was no loss of representation. No missing Republican congress critter.

Florida is supposed to be the questionable one. 21,538,187. The midpoint of the confidence interval for undercount is 3.48% which I will convert to people: 749,529. If we add that midpoint of interval to the census population, we get 22,287,716. The original count divided by the 761K gives 28.27 delegates. The adjusted count divided by that gives 29.25. So intuitively we would think Florida missed addition of a delegate, provided that we accept the estimated undercount as a true undercount. Further, if we look up Florida info, we can see they currently only have 28 delegates. So, it's good reason to believe they'd probably have 29 if they had a different census count, though we cannot say if it is valid without understanding the root cause for the discrepancy between census and sampling.

Now, apportionment is only the first step. Because the count of delegates changed in this hypothetical, they'd next have to do redistricting. We know there might be shenanigans there and so we cannot claim that the result of redistricting is necessarily fairly representative of what ought to be so far as addition of an unbiased district.

Given the extreme partisan gerrymandering in Florida, in all fairness the district ought to undo some of that...and likewise the new district ought to be reflective of changing populations in some way.

(assuming that there ought to be a new delegate).

 

[Don2 (Don1 Revised)]

So let's try this rewording:

A census does a full count of a state's population as 3,011,524. (See how this is better? The census isn't rounding to the nearest hundred thousand or estimating.) According to a survey which might be wrong and have systemic issues, an estimate of an undercount for that state is 151,781 which is a midpoint of a confidence interval.

Do you disagree so far with the changes in wording?

And what is that confidence interval,

I only found it by chance browsing over a map that suddenly had a popup with a 90% confidence interval about the %. I don't know even why they would only do a 90% confidence interval and keep it hidden. Maybe it's just because they are trying to appeal to average Joe readers? ETA: I remember seeing std error with the race "undercounts" and "overcounts" and so I think I may have also seen that with the states. So you can probably construct confidence intervals from those, provided they are there.

and what p value is used in calculating it?

No idea. That is something I searched and searched for...their threshold of significance. Convention is 5% but that would be inappropriate for 50 states and it's probably a more complex method.

While reporters will almost certainly omit such things any proper poll will have this information. Why was it not listed???

There are a lot of documents there and I haven't gone through them all. It wasn't out in front, but also, I really don't think we need it to see that the original claims were inaccurate of 6 Republican congress critters missing and 6 extra Democrat congress critters.

 

[Don2 (Don1 Revised)]

If there is an accurate enough account to show the census undercounted why don’t we use *that* count instead of the census? Am I missing something here? Sorry, I haven’t followed all the posts in this thread

Let me break down your question: "if there is an accurate enough account..."

The post-enumeration survey provides an estimate based on sampling. The estimate comes with confidence intervals. The numbers being presented are the midpoints of the confidence intervals. The "account" might not be sufficiently vetted (from my perspective) because they do not explain the root causes of discrepancies between the census count and the PES estimates. They do say that neither is perfect and that the survey estimates ought to be used as a tool for improvement in the next census. I'd only do that after verification of correctness and seeking to discover root causes.

"why don't we use that count instead of the census?"

To repeat a point I've been trying to drive home, and I know you are just checking into the thread here, ... The PES provides a range and an estimate, but the census provides a count. We can take the midpoint of the confidence interval and use that to create adjusted numbers, but there's also a chance we are changing the numbers to be less accurate, especially in consideration of sampling across 50 states with some states being small and possibly systemic issues with the sampling methodology or data.

All that said, I would expect that any sampling issues would be more likely a problem with very small states and just a few, like 2 or 3. And generally, I would agree that the numbers are probably more reflective of reality for larger states. But even those larger states' numbers may have issues due to systemic problems of unknown causes that need to be understood.

The real bottom line I think to why we do not adjust numbers is the law. We are supposed to do a census and use those numbers. The law is based not only in fairness and math but also on practicality. It took years to do the sampling and analysis and probably root out problems in the PES itself before presenting findings in early 2022. It wasn't perfect probably and they also have a deadline because of things that happen in 2023. The longer you take the better analysis you can do but the more distant from the time of the snapshot of data (April 1st, 2020), the less well you can backtrack or get correct info on residence--people's memories fail but also people lie about where they were--and the less useful the numbers become anyway because more time has transpired and your population has changed already. So you need fast, accurate results. That's the practicality side of it. The Census Bureau itself said to use the census numbers for apportionment and to use the survey estimates as a tool for future improvement.

 

[ZiprHead]

If the census estimates that a state has 2,900,000 million people, and it undercounted black and hispanic people by 200,000... what is the effect on the number of seats that state has in congress?

Zero.

You need 700,000 people to qualify for a representative...

Here is a short video that addresses this issue.

 

[Thomas II]

If the census estimates that a state has 2,900,000 million people, and it undercounted black and hispanic people by 200,000... what is the effect on the number of seats that state has in congress?

Zero.

You need 700,000 people to qualify for a representative...

Here is a short video that addresses this issue.

 

[Don2 (Don1 Revised)]

Suppose we did look at Arkansas. Its 2020 census population was 3,011,524. The midpoint of the undercount confidence interval is 151,781. So add the numbers together and you get 3,163,305. Now, divide each by 761,988. You go from about 3.95 delegates to about 4.15 delegates. Both of these easily round to 4. And if you check the apportionments for Arkansas, you will find that they do have 4 currently. There was no loss of representation. No missing Republican congress critter.

Here are some other different numbers. Another number I saw is 761,169. I am unsure how that one is derived. I tried to exclude DC population and divide by 435 but that resulted in a third number.

With respect to the figure of 151,781, it may be that the CB is taking the new estimate as the denominator when it defines the percent of "undercount" or "overcount." I have reason to think that is what they are doing. So, instead of 151,781 + state population, we'd take 3,011,524 and divide by (1 - midpoint percent in the confidence interval).

New population = 3,011,524 / (1 - .0504) ~= 3,171,361. Not that different from 3,163,305.

Dividing by 761,988... we had gotten 3.95 delegates for the census. That becomes 4.16 instead of 4.15 for the undercount delegate scenario. Both still easily round to 4.

How about if we use the other number: 761,169? We get 3.96 vs 4.17. Both still easily rounding to 4.

I did say there are other numbers floating about and this is an intuitive method. It is not going to matter for Arkansas what the variations of numbers are.

Now, apportionment is only the first step. Because the count of delegates changed in this hypothetical, they'd next have to do redistricting. We know there might be shenanigans there and so we cannot claim that the result of redistricting is necessarily fairly representative of what ought to be so far as addition of an unbiased district.

Given the extreme partisan gerrymandering in Florida, in all fairness the district ought to undo some of that...and likewise the new district ought to be reflective of changing populations in some way.

(assuming that there ought to be a new delegate).

Just want to reiterate this point. The process is:
Apportionment --> Redistricting --> Voting --> Republican or Democrat

Assigning blue or red delegates automatically at apportionment is an assumption and with changing population demographics the latter steps can be impacted producing a materially different outcome than the assumed party membership of a new delegate.

 

[ZiprHead]

If the census estimates that a state has 2,900,000 million people, and it undercounted black and hispanic people by 200,000... what is the effect on the number of seats that state has in congress?

Zero.

You need 700,000 people to qualify for a representative...

Here is a short video that addresses this issue.

While your video is funny, I don't see how it is pertinent to the subject at hand.

 

[Thomas II]

If the census estimates that a state has 2,900,000 million people, and it undercounted black and hispanic people by 200,000... what is the effect on the number of seats that state has in congress?

Zero.

You need 700,000 people to qualify for a representative...

Here is a short video that addresses this issue.

While your video is funny, I don't see how it is pertinent to the subject at hand.

It hasn't changed in 100 years...Obviously it needs to adapt, but the quality of the people involved in government needs to improve too...

Your video makes a good point.

Worst member of the House of Representatives​

The nominees:

(a) Lauren Boebert, R-Colo.
(b) Matt Gaetz, R-Fla.
(c) Ralph Norman, R-S.C.
(d) Marjorie Taylor Greene, R-Ga.
(e) Jim Jordan, R-Ohio
(f) James R. Comer, R-Ky.
(g) Mike Johnson, R-La.
(h) Thomas Massie, R-Ky.
(i) Bob Good, R-Va.
(j) Andy Biggs, R-Ariz.
(k) Scott Perry, R-Pa.
(l) Nancy Mace, R-S.C.

The winner: Too many to choose from.

 

[Loren Pechtel]

I still want to know what their measure of statistical significance is. This was based on polls vs census data--they identified the number of households polled, implying that they were random across the country. If so they simply don't have enough samples in the smaller states to be declaring these significant without setting a very low bar for significant. And if they set a low bar then this becomes simply expected noise.

Okay, so, first, if you look at the magnitudes of the "undercount" or "overcount" estimates vs the populations of the states, you DO see a correlation so that smaller states have larger "error." It has R^2 of .34 which isn't strong but isn't near 0 either. So it's conceivable there is an effect there. I was also searching for the significance and didn't find the threshold they are using...just lots of mentions of statistically significant things...

Yeah, I noticed the small state bit. And remember that smaller states mean a smaller sample in their polling, the size of those errors doesn't feel significant to me but without the details there's no way to be sure. That's why I suspect it's something like p at .10 and this is simply expected error.

 

[Loren Pechtel]

Florida is supposed to be the questionable one. 21,538,187. The midpoint of the confidence interval for undercount is 3.48% which I will convert to people: 749,529. If we add that midpoint of interval to the census population, we get 22,287,716. The original count divided by the 761K gives 28.27 delegates. The adjusted count divided by that gives 29.25. So intuitively we would think Florida missed addition of a delegate, provided that we accept the estimated undercount as a true undercount. Further, if we look up Florida info, we can see they currently only have 28 delegates. So, it's good reason to believe they'd probably have 29 if they had a different census count, though we cannot say if it is valid without understanding the root cause for the discrepancy between census and sampling.

Now, apportionment is only the first step. Because the count of delegates changed in this hypothetical, they'd next have to do redistricting. We know there might be shenanigans there and so we cannot claim that the result of redistricting is necessarily fairly representative of what ought to be so far as addition of an unbiased district.

And Florida skews elderly and red--means lots of Covid deaths. If there's an extra delegate they're probably dead.