The Volokh Conspiracy
Mostly law professors | Sometimes contrarian | Often libertarian | Always independent
Enumerated Powers and the Census Case
The Supreme Court was right to rule that the administration's rationale for adding a question about citizenship to the Census was bogus. But it would have done better to rule that inclusion of the question was beyond the scope of the federal government's enumerated powers.

Yesterday, in Department of Commerce v. New York, the Supreme Court ruled that the Trump administration's rationale for adding a question about citizenship to the 2020 census was pretextual and therefore invalid under the Administrative Procedure Act. I think this decision was correct, for the reasons Chief Justice John Roberts explains in his majority opinion. But the Court would have done better to simply rule that the inclusion of the question on the census was outside the scope of federal power under the Enumeration Clause of Article I of the Constitution. The majority's reasons for rejecting that argument strike me as weak.
Article I Section 2 of the Constitution requires Congress to make "an actual enumeration" of the population every "ten years," for purposes of apportioning representation in the House of Representatives. The enumeration is to be done "in such Manner as [Congress] shall by Law direct." This latter phrase undoubted gives Congress broad power to determine exactly how the census shall be done. But broad power is not unlimited power.
Congress clearly has the authority to include almost any questions on the census that would increase the accuracy of the resulting count. It also has the power to ask a wide range of questions that would not have a significant effect on accuracy either way (especially if those questions involve information-gathering of a type authorized by Congress' other enumerated powers, which would provide additional authorization for them).
But matters are different if the addition of a given question is likely to reduce the accuracy of the count. Some of the powers the Constitution grants to the federal government are powers to regulate a particular type of activity, largely without regard to the purposes for which the regulation is undertaken. For example, Congress' power to regulate interstate commerce can be used for a wide range of different goals, some of which may even conflict with each other.
By contrast, the Enumeration Clause is a power to achieve a particular type of objective: in this case an accurate head count of the population of the United States. Indeed the Enumeration Clause is a rare constitutional provision that not only grants authority to pursue a particular objective, but actually mandates it. The Clause requires the government to conduct a new enumeration every ten years. The power in question therefore cannot be used to authorize policies that actually impede its mandated purpose. And that is exactly what the addition of the citizenship question would do.
Survey research experts, including the federal government's own experts at the Census Bureau, agree that including the citizenship question would substantially reduce the accuracy of the count. The reason is that a significant number of immigrant and other households that include non-citizens would likely choose not to participate, for fear that they or their relatives might end up getting targeted for deportation as a result. Even if people in question ultimately forestall deportation by prevailing in court, a legal battle of this type can be painful, costly, and disruptive. Many would prefer to avoid even a small chance of inviting it.
While the Enumeration Clause gives Congress the power to choose the "manner" of conducting an enumeration, that is not the same thing as using that power to undermine the very goal it is supposed to achieve. The power to do X is not also a power to impede X. If Congress and the president had unlimited authority to conduct the census in any way they want, unconstrained by the requirement that the methods must further enumeration rather than impede it, they could, for example, adopt a law under which anyone counted by the census would have to pay a tax in order to be included. This would clearly be a regulation of the "manner" of enumeration. But, just as clearly, it would be unconstitutional, because it would lead to widespread undercounting by creating an incentive to avoid being counted.
The Trump administration's proposed citizenship question is just a less extreme example of the same problem. Indeed, the administration likely added the question precisely because it would reduce the accuracy of the count in states with large immigrant populations, thereby reducing the number of congressional seats allocated to those states. But even if the administration's motives were pure, the addition of this question still undermines enumeration rather than furthers it, and thereby falls outside the scope of the federal government's Enumeration Clause powers.
A ruling against the question based on the Enumeration Clause would have ended the case immediately (as opposed to the current remand for further consideration of possible alternative rationales for including the citizenship question). It also would have obviated the need for consideration of the administration's motives and the procedure by which it decided to include the question.
Chief Justice Roberts' best argument for ruling that the citizenship question falls within Congress' authority is that a similar question was included in all but one census between 1820 and 1950, and the question was also given to sub-samples in several censuses since then. But this evidence is less impressive than it might initially seem to be.
During the 1820-1950 period, we did not have social science evidence showing that inclusion of a citizenship question would reduce accuracy. While the meaning of the enumeration power was the same, whether the inclusion of a given question furthers the authorized purposes of that power or impedes it is a factual question. This is one of a number of situations where originalists and textualists can and should take account of new factual evidence. The resolution of any legal case depends on a combination of textual meaning (which, at least under originalist assumptions, does not change) and factual evidence (which can indeed change, and often does).
Moreover, during most of the time from 1820 to 1950 there were few or no federal immigration restrictions, and even less in the way of systematic federal efforts to deport undocumented immigrants from the interior of the country. Thus, there was little incentive to avoid participation in the census for fear of deportation. The situation today is, for obvious reasons, very different.
Roberts also notes that the census has always been used to ask a variety of other demographic questions, such as ones about respondents' age and sex. But including these questions creates little or no risk of creating a significant undercount.
Assessing whether the inclusion of a census question undermines the goal of enumeration would require courts to consider social science evidence. But no more so than is routinely done in a wide range of other routinely heard by federal courts, such as antitrust cases and redistricting cases where plaintiffs allege racial discrimination. It might be reasonable to defer to the federal government's judgment in close cases, where the evidence is ambiguous. But not so in cases like the citizenship question, where it is overwhelmingly on one side.
As both Chief Justice Roberts and Justice Stephen Breyer (in his partial concurring opinion), note, if the federal government cannot include a question on citizenship in the 2020 census, it can still do so in other surveys. Census Bureau experts recommended exactly that course of action. It is likely that a survey of citizens can be authorized under one of Congress' other enumerated powers (especially under the very broad modern interpretations of some of them), even if not under the Enumeration Clause. But the federal government cannot include such a question in the census mandated by the Enumeration Clause.
This flawed part of yesterday's decision probably will have little effect on cases outside the context of the census. In that sense, the harm it causes will be limited. But it is still unfortunate that the five conservative justices (all of whom joined this part of the the Chief Justice's opinion) signed on to such a seriously flawed ruling which allows a federal power to be used for the opposite of its textually required purpose. That goes double for those, like Clarence Thomas, who are generally very careful to enforce structural constraints on federal power in other contexts.
The four liberal justices also deserve some criticism here. They all refused to sign on to this part of Roberts' opinion, which suggests they likely disagree with it. Yet none bothered to explain the grounds of that disagreement. It would have been helpful if they had done so.
UPDATE: It is worth noting that non-citizens (and other non-voters) have always been included in the "enumeration" required by the Enumeration Clause for apportionment purposes. Indeed, such inclusion is mandated by the text of the Constitution, which mandates a count of all "free persons," except for "Indians not taxed."
Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Comments may only be edited within 5 minutes of posting. Report abuses.
Please
to post comments
I beg to differ, completely: https://realitiesblog.com/2019/06/27/the-citizenship-question/
That's a pretty expansive reading of the enumeration clause, in any event that reading would be superseded by section 2 of the 14th amendment, which requires knowing the "whole number of male citizens twenty-one years of age in such State." The only way to definitively know that number of course would be by the census asking whether the respondents are citizens.
Thanks for telling Somin about the other parts of the Constitution.
Somebody had to.
No. It's perfectly possible to do a different survey to find that out, like the ACS, or to rely on administrative records which will be more accurate than asking the question on the census.
Nothing in the 14A requires that the information be obtained specifically through the census.
Yes, and equally, nothing requires that the information specifically NOT be obtained through the census.
Yes, but Kazinski is claiming that the census is the only permissible way to get the data.
IOW, he claims that the requirement to get the data implies that the census must ask the question. That's not right, IMO.
The requirement to get the data implies that the question must be asked. At that point, the argument that it can't be asked in the census collapses totally, amounts to a demand that we conduct TWO censuses every 10 years, one with, and one without, the question.
No. It doesn't.
It requires that the data be obtained. It is obtained.
Indeed, I'd argue the opposite. The implied requirement is that the most accurate method of obtaining it be used. That's not by putting it on the census, as the people who actually understand this stuff have uniformly said.
The problem (as with Prof. Somin's argument) is that there is no such implied requirement. There can't possibly be. If we doubled — or tripled — the Census Bureau's budget, don't you think it could do a lot more to ensure that the census was accurate? Send out twice as many people to do followup interviews of nonrespondents? Hire a whole staff of people to double check whether those who did respond did so accurately? Etc. Does that mean that it's unconstitutional not to double or triple (or quadruple, if that's what it takes) the census bureau's budget?
Why would that be more accurate than the census? IE, actually counting the people?
Because including the question in the census will distort the results. The answers won't be accurate.
But sampling will be more accurate?
On the whole it will be as accurate, and any errors won't affect the census count, which of course has a pretty important purpose besides data-gathering. Remember, we are talking absolutely huge samples here.
One way a sample is actually better, as I understand it, is that you know how many respondents there are, so you know the non-response rate. Not so in a census.
Bernard, can you say more about suggested census sampling methods, so I can get a better ability to understand your last paragraph. Seems like there would be at least two kinds of non-response. One kind would be someone you know about, who doesn't answer. The other kind would be someone you don't know about and never find.
Not seeing how sampling picks up that latter kind, even as a non-response. I don't think the census always picks up that kind either, of course. It might do better if it want back to universal door-to-door census takers.
The idea of using sampling to supplement the census has been around for decades - it was being shot down yet again when I (briefly) worked for the Census Bureau in the 90s.
The most common concept works like this: non-answer rates vary by contact method. If you compare how many people you contact by mail with the number you contact by phone or in person, you can use the rates to fill out some charts that give you an estimate of the number of people that failed to respond to ALL methods.
There are several problems with this, starting with the fact that non-answers are NOT random (a strict requirement of the math). In fact, when these methods were tested, they were consistently wrong.
That's not how I understood sampling to work - it wasn't a way to address nonresponses, but rather to concentrate your efforts on a random sample to attain higher response rates there, and then extrapolate based on that.
That would be more accurate than the current style of attempting exhaustively counting every person.
The Census requires that we count everyone in the group. That's a definition and a base requirement. There has never been a (serious) proposal to stop doing that.
No, the sampling proposals are attempts to deal with the problem of non-answers: the undercounts. That group is the only group that we are concerned with, because every other group has been counted already.
And no, a sample cannot be more accurate than a census. To put it mildly, any sample would have the same problems as a incomplete census, but with a smaller sample size - and thus higher error.
non-answers are NOT random
This is a major problem with the citizenship question. If you ask a question that reduces response rates randomly you get an inaccurate total, but the proportions should be right.
With the citizenship question that's not so.
This has long been my understanding. People have considered using sampling to supplement the census, but never replace it, because actual enumeration is superior. Much more labor intensive, but superior.
It's not that actual enumeration is necessarily superior; it's that actual enumeration is required.
(They do use sampling to supplement the census — but not for apportionment purposes.)
Stephen,
I know they use some complex sampling methods. I don't know the details. Maybe Toranth does.
If you mail out 100,000 questionnaires and get 90,000 responses, you know 10,000 are missing.
Toranth,
I thought sampling for the census was abandoned because it was decided, by someone, that the requirement for an enumeration precluded using it, not because it would be inaccurate. I remember that there were claims that it would be more accurate than an enumeration.
Sampling as a replacement for the Census would have been unconstitutional, correct. Sampling was a proposal that would be used to supplement the Census - by attempting to estimate the number of undercounted.
If the non-answers were random, and a few other assumptions about independence and distributions were satisfied, then YES a sampling estimate could be used to estimate the number of uncounted respondents, within a known reasonable estimate of a margin of error.
The problem with assuming the mailing 100,000 questionnaires and assuming that you should get back 100,000 responses is that you are probably not accurately targeting your population. A non response could mean that the address is unoccupied, or doesn't exist any more. New addresses might not be in your list, and be missed by not sending a questionnaire. A single address may have have multiple resident groups, that should have received multiple questionnaires.
That's why the Census uses multiple methods to contact respondents - because any single one is almost certain to be incomplete.
Indeed. In fact, the only case where I would use sampling for the Census, is in determining which areas and populations may need extra resources in obtaining responses.
You're talking about a differrent sampling than everyone else is.
Also, if you don't understand why sampling would be more accurate than enumeration, you either don't understand statistics or don't live in the real world.
Sarcasto, please pay attention.
Once more - a sample CANNOT be more accurate than an enumeration. It is not possible. Period.
An enumeration has zero MSE.
A sample will have an MSE >=0. Never less than zero.
Therefore, a sample will never be more accurate than enumeration.
Toranth, YOU pay attention:
You're making an argument from theory. In the real world, where not everyone responds, sampling is more accurate than enumeration because response rate is related to measuring effort.
Again, stop.
You don't understand this: You CANNOT come up with any method that will produce a more accurate answer than an attempt to count all the things.
YOU CANNOT DO IT.
If you somehow came up with a method to get more answers from people in a sample, you would just use that same method to get the answers in your enumeration - thus improving the enumeration over the sample again.
Sarcastro, let's use a demonstration here, on why a census is more accurate than sampling.
Let's take a forest, that's 100 square miles. A census would individually count each individual tree, and would come up with a number, say 158,295 trees. It would be tedious and expensive, but it would give an accurate count
Sampling, by contrast, would only count the trees in a given section, then multiply it to the rest of the forest. It would count a square mile of trees, and see 135 trees in that section. Then it would say there are 135,000 trees in the forest. A significant undercount.
See?
Are you both misunderstanding me on purpose?
People aren't trees; some of them will be missed in any count.
This causes errors in both methods.
However, due to the smaller sample in sampling it is easier to use various methods to counteract this effect. With enumeration, these methods are not available.
Thus, while in theory-land completely successful enumeration is superior to completely successful sampling, in the real world sampling will provide a more accurate count.
Got it now?
Edit. 1350 trees in a square mile.
Your argument here is wrong because it begs the question. Of course accurately counting every single item is more accurate than picking a random sample and extrapolating.
But what if you individually count only 132,621 individual trees, because 25,674 of the trees you were trying to individually count hid when you came by to count them? (Okay, the analogy breaks down a bit there. Perhaps they were in an inaccessible area of the forest and so you couldn't get to that region to count them.)
Ah, now we are getting somewhere.
This sentence is key.
"Of course accurately counting every single item is more accurate than picking a random sample and extrapolating."
Which is the major point being made by myself and Toranth. Please keep this in mind.
Actual enumeration will always be more accurate in theory than sampling and extrapolation.
Now what you and Sarcastro are arguing about is application. Now sampling has many of the same problems that enumeration does, because sampling involves counting, but a smaller sample. So, sampling can run into the same issues of non-response bias, etc. However, assuming the same methods and resources per subject being counted, once again, enumeration will be more accurate, as it avoids sampling bias. Both methods (Sampling and enumeration) are subject to non-sampling error. Of course, at the same number of resources per subject, a census will be much more expensive.
Now, in order for sampling to be superior to enumeration, in practice, the non-sampling error rate in enumeration would need to exceed the combination of the sampling and non-sampling error rate in a sampling study. One issue is, any methods taken to reduce the non-sampling error in sampling could be extended to enumeration.
Now, in terms of the census being inaccurate by use of actual enumeration, thus requiring sampling, the issue is, the argument is a bit circular.
Because sampling is used to provide the "real" population that the enumeration "missed". So, it's a circular argument. We need sampling to fix the numbers because sampling showed the numbers aren't quite accurate.
AL, I'm not buying you've cleverly at least bracketed DMN and I into explaining how we're not talking about theory since I've been saying exactly that for 2 days as you continued to misunderstand.
As to your latest argument, that the methods of undercount reduction in sampling can be scaled to enumeration, that's not much more of an understanding.
What can be done for say 100K would be cost prohibitive for the entire US population. Come on, man.
sampling is used to provide the “real” population that the enumeration “missed"
As I've said before, this is not what anyone else is talking about when they talk about sampling.
Sigh,
Sarcastro. You do realize the census has an enormous budget, in the billions of dollars, with literally hundreds of thousands of temporary employees, right?
Furthermore, how exactly do you propose using a sampling of 100,000 people to obtain an accurate total population of each of the 50 states (not to mention PR, Guam and the other territories). You can't just call random people. You need to do it based on area. But the population density is extremely heterogenous, which makes it very bad for sampling. And the random "areas" you could pick (Zip codes, census tracts, census blocks) vary enormously in population.
Go on...think. Pick a method. I'm listening.
Sarcastro, you honestly do not seem to understand any statistics at all.
First off, a sample does not produce a population count. It produces an estimate of the population count within a margin of error. That error CANNOT be ignored, and cannot go away.
Second, as AL and I have repeatedly tried to explain, any method you can use to "improve" the response rate of a sample can also be used to improve an enumeration. There is no possible other way it can be - an enumeration is nothing but a sample where sample size = population size.
Next, you claimed that you aren't talking about samples they way we are. In which case, you need to explain exactly what you mean. Because if you aren't talking about sampling estimates to supplement the Census and you aren't talking about sampling to replace the Census, you haven't even implied what you are talking about.
So: Explain exactly what your sampling would be, and what you intend to estimate. Then explain your methodology and how it can be used for a small sample but not a enumeration.
I'll accept links to research papers, if they are publicly accessible.
In other words - put up, or shut up.
re: "If you mail out 100,000 questionnaires and get 90,000 responses, you know 10,000 are missing."
No, you don't. Or more precisely, you do know that if and only if you have prior knowledge that every one of your 100,000 destinations was valid. That is not and never will be true.
Consider just a few counter-examples:
- One of your questionnaires goes to a home that is currently vacant. This does not represent an undercount - this is a legitimate zero result.
- Another questionnaire goes to an address that doesn't actually exist due to a typo in your database.
- A third goes to a home that you believe to be a single-family dwelling but is actually a duplex. The recipient recognizes the mistake and makes a copy for his neighbor to fill out.
- You fail to send a questionnaire to the guy living on a cot in the backroom because you only sent questionnaires to residentially-zoned areas.
Somin,
Your "analysis" of enumerated powers would also remove questions about race, national origin, and computers and smart phones in the household.
Only a body count would be allowed.
If those metrics are proven to be a disincentive to getting a good count, then yeah, kill 'em.
Really we should go to sampling for maximum accuracy, but that's not what the Constitution allows.
Let’s try to make this as painless as possible. One of the kinds of taxes the Constitution expressly permits Congress is a capatation tax, based expressly on a census or enumeration.
Another of Congress’ express powers is a treaty power that has long been interpreted to permit Congress to decide how foreign citizens should be taxed, including exempting them from taxes.
So Congress’ enumerated powers include the power to enact a capitation tax, based on a census, that applies to citizens only.
Given this, it necessarily follows that it can provide for a census that contains a citizenship question.
And this is all based on an extremely narrow, pre-20th-century view of Congress’ enumerated powers that doesn’t even touch the Necessary and Proper Clause.
Well, I see one tiny flaw here. The Census is not supposed to be used to identify specific individuals, so it wouldn't provide the information as to whether a given individual is a citizen.
In addition, if it were used this way, there would be a powerful incentive to lie.
Besides, it's easy enough to impose your fantasy tax universally and give non-citizens the ability to claim an exemption by demonstrating that they are not citizens.
The purpose could be to apportion the tax among the several states based on proportion of citizens.
But more fundamentally, for your statement “the census is not supposed to be used to identify specific individuals,” this is a relatively recent statutory policy. The earliest censuses recorded names. And Congress need merely pass a law if it wants to restore the old approach. Nothing in the constitution prevents it.
Note the Wikipedia entry on the Census of 1800, which lists the question. The first question asks for names.
https://en.m.wikipedia.org/wiki/1800_United_States_Census
Except that identifying individuals would in all likelihood significantly reduce response rates.
The purpose could be to apportion the tax among the several states based on proportion of citizens.
I don't get your point. Why wouldn't any method of collecting the "citizen tax" give you the data?
Really? I've looked up my great-grandparents, and grandparents census records. They looked personally identifiable to me.
There have been various statutes concerning confidentiality. The most recent was in 1954, I think, probably later than the records you saw.
"I think this decision was correct, for the reasons Chief Justice John Roberts explains in his majority opinion."
Now we know for sure that Roberts was wrong.
"Indeed, such inclusion is mandated by the text of the Constitution, which mandates a count of all "free persons," except for "Indians not taxed.""
The actual text reads, "Representatives and direct Taxes shall be apportioned among the several States which may be included within this Union, according to their respective Numbers, which shall be determined by adding to the whole Number of free Persons, including those bound to Service for a Term of Years, and excluding Indians not taxed, three fifths of all other Persons. "
This actually required enumerating ALL persons, as the 3/5ths clause required knowing the number of unfree people, too.
But Kazinski is right: Implementing Section 2 of the 14th amendment would actually require knowing the number of citizens, not just people.
Anyway, didn't Roberts implicitly address your argument, by pointing out that the contrary position would declare almost every Census ever conducted to have been unconstitutional? Which is a rather radical conclusion, but follows directly from your reasoning.
In any event, and I do mean this seriously: The only group anyone might reasonably expect to under-report as a result of this question would be illegal immigrants, who are already subject to deportation. Why couldn't we remedy any under count by simply deporting enough illegal aliens to make the under count reality?
Indeed, if you're actively deporting illegals, an under-count of them could actually make the Census more, not less, accurate.
Brett,
Don't get Somin to draft your brief to SCOTUS.
Um, Brett. Some households have both legal and illegal immigrants.
Also, do you really think it's a good idea, or even feasible, to organize ICE deportation forces - immigration einsatzgruppen - to sweep up hundreds of thousands, millions, of people and deport them?
I know Trumpists don't consider illegal immigrants, or perfectly legal asylum seekers, to be fully human, but they are wrong about that.
True on the first point.
Yes, I actually do think it is both a good idea, and feasible, to organize ICE deportation forces to sweep up hundreds of thousands, millions of people, and deport them. Though I think at the hundreds of thousands point, most of the rest would self-deport.
This has nothing to do with whether they're considered fully human. They're fully human people who legally have no right to be in the US. Expelling them undoes an ongoing crime, proving that possession is not, in fact, 9/10ths of the law, and that refusal to enforce a popular law can't establish facts on the ground that will never be undone.
Oh, and the "Einsatzgruppen" weren't deporting people who had illegally entered a country. They were death squads rounding up and killing people who were citizens of those countries. Democrats are already doubling down on the 'concentration camp' nonsense, going for triple?
I actually do think it is both a good idea, and feasible, to organize ICE deportation forces to sweep up hundreds of thousands, millions of people, and deport them.
You reveal yourself a monster.
The concentration camp business is not actually nonsense. There were Nazi concentration camps whose purpose was simply to detain undesirable individuals under brutal circumstances.
Not much different than what Trump is doing. Are they going to kill people? No, but it's pretty clear that they don't care if they die. Pouring out water left in the desert, arresting those who provide it, are not the behavior of an organization that much gives a damn about people's lives.
And you reveal the left-wing disease: Anybody who disagrees with you is a monster. Well, remember what Friedrich Nietzsche had to say on the subject:
"Whoever fights monsters should see to it that in the process he does not become a monster."
I suspect the reason the left is so quick to declare those who dissent from its policies "monsters", is that it frees them to be monstrous, and feel good about it.
Naw, Brett - I'm a bit more sanguine about the camps, but it's not a liberal thing to have moral lines when it comes to caring about the needy.
Or, at least, it didn't used to be.
I'm not proposing any camps, I think we can go direct to returning illegal aliens to their country of origin.
And this isn't about caring about the needy. It's about continuing your generation long effort to, in Bertolt Brecht's words, "elect a new people".
Look at what's going on with Democratic policy on illegals: It's progressed from, can't do anything to stop their entry, to giving them drivers' licenses, to free college educations, to medical care.
You're clearly not content with their coming here illegally to take part in our economy, you're now proposing to bribe them to come here.
Given your attitude towards gerrymandering, voter suppression, and the census issue to complain about Democrats trying to elect a new people is ill-advised, IMO.
It's the Republicans who are trying desperately to disenfranchise likely Democratic voters.
And they have the help of SCOTUS, and its RNC wing.
We wouldn't have to do those things had liberals not flung our borders open to the third world in 1965. If you only counted the votes of the population that was present in 1965 and their descendants, the Republicans win over 70%.
Brett, it's the GOP that is trying to elect a new people. A smaller, older, whiter, more rural one than the one actually there.
They're pretty open about it. Roberts and the other four Republican operatives on the court are doing all they can to help.
Except that the "one actually here" is one that was foisted onto an unwilling population. Americans did not want these people. We still don't.
No, Brett, anyone who disagrees with me is not a monster. But here's the point.
There are illegal immigrants, lots of them, who have been here for years. They work, pay taxes, etc., and have some sort of roots in the US. Their children, in most cases, have no connection with Mexico, or Honduras, or wherever.
Now you are advocating sweeping all these people up and sending them back to wherever, for no reason, ultimately, other than that we can. Because that's what your argument about, "they're illegal" amounts to.
That's monstrous. It's a terrible way to treat people, immoral, inhumane, cruel.
If you disagree on this specific issue then, yes, you are a monster, and I see no way in which I, and those who agree with me, are becoming monsters.
Why have immigration laws at all?
Meth manufacturers pay taxes, too. Should we leave them (and their children) alone?
How can enforcing a law that's been around for decades be "monstrous "...?
Yes.
It's hard to see how the mere passage of time would make a monstrous policy not monstrous. (Jim Crow laws were "around for decades.")
Except most of them have genetically low IQs, are way overrepresented in special education classes, gang membership and criminality as they get older. There's a reason that Honduras and Mexico are shitholes. Because they are composed of genetically inferior Aztecs and Mayans who are incapable of creating civilization. And that's also the reason that liberals hate Cubans. Because most of them are white and successful.
Oh shut up with your idiot "genetics." Your IQ is dismal, whether for genetic reasons or otherwise.
And by the way, it seems that immigrants, even illegal immigrants, have lower crime rates than native-born citizens. Not that I expect you to abandon your bigotry.
Truth hurt? And no, it doesn't "seem" that. Hispanic immigrants have a crime rate far in excess of the white population. Yes, it's possible that illegal Hispanics have a lower crime rate than the population as a whole when you include Hispanic and black citizens, but that inclusion is a dishonest metric.
ICE acted exactly the same under Obama, no Nazi comparisons there.
You just run in the wrong circles - it was there.
Oh really? Rounding up hundreds of thousands, as Brett would like? I missed that.
But so what? Is it your claim that anything Obama did is OK?
His claim is that it was equally wrong then, but that the people screaming the loudest were silent then. He’s also alluding to a common observation that this isn’t the only example of such, and that every example has the same political valence - it’s bad if your guy does exactly the same thing I ignored when my guy did it.
I doubt he thinks it was wrong.
Anyway, the fact is that Obama's deportation policies were widely criticized, though it's worth noting that he deemphasized "interior" deportations - the deportation of those in the country with roots in their community.
He also established DACA, which the Trumpists hate, for reasons that are not admirable, IMO.
"He also established DACA, which the Trumpists hate, for reasons that are not admirable, IMO."
Only if you assume the worst. I'm assuming you don't approve of Trump moving funds to pay for his wall given that congress declined to appropriate new money for that purpose. In the scheme of things DACA was more far-reaching in it's implications and that is why I hated it.
"But the Court would have done better to simply rule that the inclusion of the question on the census was outside the scope of federal power under the Enumeration Clause of Article I of the Constitution. "
They won't go there because that would invalidate the entire long form of the census. I wish they would, but they won't.
I don't know why it makes earlier censuses unconstitutional. As Ilya points out, the facts are different now. Besides, what does it matter?
The 1970 long form is declared unconstitutional. BFD.
I don't see how any facts alter the issue of extra census questions being outside Congresses enumerated powers.
First, of course, you sneak in the assumption that previous censuses were Constitutional. Maybe not. so what?
But say they were. Then the presence of large numbers of illegal immigrants will have a greater impact on response now than before.
Most of the time in the past the question wouldn't have much effect. There weren't enough illegal immigrants to worry about.
Further, it may well be that we have learned enough to appreciate the danger. If it was a problem in 1950, say, we were unaware of it. Today we are.
"First, of course, you sneak in the assumption that previous censuses were Constitutional."
I make no such assumption.
I said the court wouldn't go there, because that's a necessary result of invalidating ANY census long form question on enumerated power grounds.
That is not in any way shape or form ME making any assumption about the constitutionality of past censuses. In point of fact I specifically said I wished that the court would go there even though I know they never will.
"But say they were. Then the presence of..."
None of those facts has any relevance to whether or not questions beyond the minimum needed to generate the enumeration required by the census clause fall outside of Congress' enumerated powers.
I said the court wouldn’t go there, because that’s a necessary result of invalidating ANY census long form question on enumerated power grounds.
Well, the court might want to go there, but I don't see the harm if they do. The 1970 long-form is declared unconstitutional. So what? Do respondents sue to get the data back? Or have it erased from government files?
Like others, I would like to know how Professor Somin's view is affected by U.S. Const. art XIV, sec. 2 ("But when the right to vote ... is denied to any of the male inhabitants of such state ... or in any way abridged, except for participation in rebellion, or other crime, the basis of representation therein shall be reduced in the proportion which the number of such male citizens shall bear to the whole number of male citizens twenty-one years of age in such state."). I don't think this is as slam-dunk of an argument as others think it is -- maybe the predicate condition can never occur because of the fifteenth amendment -- but it seems odd to omit discussion of it.
The 15th amendment quite simply says, "The right of citizens of the United States to vote shall not be denied or abridged by the United States or by any State on account of race, color, or previous condition of servitude." That's hardly an exhaustive list of possible grounds for denying the franchise.
While Section 2 says, "except for participation in rebellion, or other crime".
Not to draw a Venn diagram, but it doesn't take much imagination to come up with a basis for abridging the right to vote on a basis which isn't "race, color, or previous condition of servitude", but neither is "participation in rebellion, or other crime.
You could, for instance, condition the franchise on the basis of education, denying it to anybody who doesn't have at least a high school equivalent. This wouldn't be prohibited by the 15th amendment, but would still invoke section 2 of the 14th amendment.
So, no, the 15th amendment didn't obsolete Section 2.
Seems like quite a bit of hand waving to pretend a personal goal is a constitutional one.
If the question was constitutional before, and the constitution has not changed, then the opinions of "social science" people don't make it unconstitutional now.
Why even have a Constitution if you want judges to ignore or downgrade or upgrade bits of it arbitrarily to fit your personal preference? Just say to want to rule by decree. It's simpler and more honest.
If the question was constitutional before, and the constitution has not changed, then the opinions of “social science” people don’t make it unconstitutional now.
Of course they can. Isn't it possible that circumsatnces have changed, that facts are different now than they were then, that social scientists - statisticians in this case - actually have useful knowledge not previously available.
I know many lawyers, including notably Roberts, quail at the sight of a number, but that's a poor excuse.
Real statisticians are mathematicians not social scientists.
But if you are talking about people who use just use algorithms and applications to serve their bureaucratic political or sociologist masters. Then they are clericals or technicians.
political or sociologist masters
The shadowy sociologist cabal revealed!
Seriously, though, sociologists so some pretty beefy stats stuff. Myself from 5 years ago can't believe I'm writing this, but just because it's applied doesn't mean it's not robust.
Most particle physicists use complicated statistics frequently, but they're not statisticians. The best medical statistics experts might properly be call statisticians, but no one would call them "social scientists" or mere technicians.
HEP folks use data processing more than they do stats.
But having gone to some social sciences conferences, they get pretty deep into some cutting edge statistical analysis. Maybe too deep...
Real statisticians are mathematicians not social scientists.
No. This is just plain wrong. You have no idea what you are talking about.
No, he's 100% accurate. I taught statistics to "social science" students. I worked as a professional statistician.
In the physical sciences, they say "The plural of anecdote in not data".
In the social sciences, they say "The plural on anecdote IS our data".
In psychology, they say "What's a 'plural'?"
Coming from the physical sciences, I was uncomfortable when I first encountered social science as well, also due to their style of in-depth study of a single one-time situation.
But upon asking around, part of the amelioration of this seeming problem is that their claims are similarly much more modest than the physical sciences - it's a descriptive mode much more than it's predictive. Any predictive aspects are much more on the margins, and come from concatenating many dissimilar studies that share similar margins.
Dunno about psychology - never had any experience with those folks. But I do understand that they are no less accurate than the supposedly more clinical psychology.
Social sciences suffer from an extreme vulnerability to methodology fiddling and small sample sizes.
The small sample sizes mean your design needs to be very precise - a near impossible requirement when you don't understand the population in the first place.
The methodology issue is that answers given in surveys or interviews are very sensitive to the design. Even minor phrasing changes can produce different results - as demonstrated in this classic documentary.
This is why the social sciences produce "positive" results with p-values that would make physical sciences die of laughter - <0.2 is not an uncommon value.
Not to mention p-hacking, which haunts all of modern science.
Social science is aware of that issue, and every project I've seen has a section devoted to addressing the wording issue.
And if you think sample size is an issue, you're not up on the latest social science. Thanks to social media, that's not really a thing anymore.
Sarcastr0, you might be interested to take a look at neuropsychology. Not sure it is all they do, but one thing neuropsychologists do is give folks batteries of tests to probe brain function. After a bit of experience watching that testing at work, I was astonished by the repeatability of the results, over time, and from examiner to examiner. What made it more impressive was that the aim was not to quantify physically measurable variables, but instead to characterize fairly abstract notions, such as executive function, or spatial orientation ability. And it turns out that results thus measured do have useful predictive implications.
Real statisticians are statisticians. Many work in the social sciences, doing things like experimental design, including survey design. You know, the kinds of things the census needs.
They are not working with "opinions." They work with mathematically solid tools.
"Of course they can. Isn’t it possible that circumsatnces have changed, that facts are different now than they were then"
Maybe, maybe not. It depends on what constitutional question you are asking. Certainly all that would be relevant to a challenge on equal protection grounds. However,on the enumerated powers question, Somin's stated preference for invalidation reason, those changes in facts/knowledge are of precisely zero relevance.
If the wording of the question is the same, then facts haven't changed. More knowledge about facts doesn't change the facts. They were always there, regardless of anyone's knowledge of them.
Social scientists also don't know facts about events that have not occurred. They may have opinions about what may occur, but those are not facts.
There is a huge difference between random "opinions" and statistical projections.
Just ask casino owners if they design their games based on "opinions."
I agree that statistical projections can be very accurate, but they are still not the same thing as facts. Your casino operator might know that on average blackjack players are going to lose 5% (or whatever; I made the number up) to the house, but that doesn't mean that John Smith playing blackjack tonight will actually lose 5%, or even that all the people playing blackjack tonight will actually lose 5%.
From the discussion, I think highlighting this from the OP would be helpful:
"During the 1820-1950 period, we did not have social science evidence showing that inclusion of a citizenship question would reduce accuracy. While the meaning of the enumeration power was the same, whether the inclusion of a given question furthers the authorized purposes of that power or impedes it is a factual question. This is one of a number of situations where originalists and textualists can and should take account of new factual evidence. The resolution of any legal case depends on a combination of textual meaning (which, at least under originalist assumptions, does not change) and factual evidence (which can indeed change, and often does).
Moreover, during most of the time from 1820 to 1950 there were few or no federal immigration restrictions, and even less in the way of systematic federal efforts to deport undocumented immigrants from the interior of the country. Thus, there was little incentive to avoid participation in the census for fear of deportation. The situation today is, for obvious reasons, very different."
Just so long as people don't keep saying 'What Somin neglects to address is...' and then discuss exactly what he addresses here.
[…] from Law https://reason.com/2019/06/28/enumerated-powers-and-the-census-case/ […]
If I were alive in 1787, I might construe the Census Clause very narrowly, to permit merely the population counts necessary to resolve legal issues, and some power under the Necessary and Proper Clause to ask additional questions to ensure the accuracy of the count.
But we've had over 200 years of continuous practice where the Census Bureau can use its mandate to engage in general population demography. And that counts for a lot in a common law system. So the Court's holding that the Bureau has the power to ask this seems right to me.
All this talk about the enumerated powers and still no one has invoked the term "air force"...?
It seems that ANY question beyond simply counting the population is beyond any enumerated power specified in Article I, section 2. So, if any other question is acceptable to ask (number of bathrooms?) then any other question must be acceptable to ask (are you a citizen of the United States?)
Why?
Asking about bathrooms is not going to distort results.
Whether or not a given "extra" census question would distort the results, might be relevant to other constitutionality questions, but it has no relevance to the enumerated powers question.
Congress either has the power to add extra questions to the census or they don't. Neither the nature nor the supposed effects of the question(s) are relevant.
Matthew, I'm not following that. Why wouldn't a question which did not distort the objective of the census be okay? And why would a question which did distort census accuracy be okay? The enumerated power here is to get an accurate count, and it is mandatory.
No; the enumerated power is to ask questions. "Get an accurate result" can't be a power.
A few points.
1. Knowledge of the number of citizens in a given area is a constitutional requirement. Race, renting or owning, is not. Gender and age is.
2. The Census accurately collects this information, and has for hundreds of years. Sampling can be prone to systematic bias. Actual enumeration provides, at a bare minimum, a critical baseline.
3. Ilya misses the mass deportations in the 1920's and 1930's known as the Mexican Repatriation, where hundreds of thousands to millions of Mexicans were "repatriated". And yes there was a citizenship-type question on that census in 1930.
Sampling done correctly will be as accurate as actual enumeration, maybe more so. Stop cutting and pasting a talking point from some moron.
1. "Done correctly." It may not necessarily be done correctly.
2. If you only do sampling, and not an actual enumeration, it will be impossible to tell if it's actually done correctly.
Keep in mind, any sampling has many of the same implicit issues as an actual ennumeration, including lack of response. The census attempts to get around this, by going to significant lengths to obtain 100% response rates. Sampling doens't, and just attempts to "guess" at the right answer, which can magnify error rates.
When we vote, there's a reason we count the actual votes, rather than just "Sampling" or "polling". There's a reason for that. If you can actually count every individual, that will be superior to a sampling method. It's only when you can't count every individual, that issues may be apparent.
Wow, you are really showing off your ignorance today.
No, sampling will NEVER be more accurate than an actual enumeration. It is physically and mathematically impossible for that to be true.
Period.
Properly done sampling is going to be more accurate than properly done enumeration because you can mitigate undercounts by using more exhaustive techniques when you're not dealing with the entire sample.
I'm no statistician, but that's like the first day of stats 101.
As for AL, if you think sampling could be done badly I dunno why you think enumeration would be any better. And if you don't see why voting is different from counting...
No, a properly done enumeration is a complete count. This is by definition.
You cannot be more accurate than a complete count. Period.
Properly done counting (enumeration), gives a 100% correct answer 100% of the time, because it counts 100% of the item desired to be counted.
Sampling and stats are a way to minimize the workload, and measure the error in that minimized workload. Enumeration doesn't have multiple different methods, as sampling does, which can give several different answers. To enumerate, you literally count. It's pre-school level math.
You can't do proper enumeration in the real world. You can't do proper sampling either.
But the second has a better failure mode than the first.
No, it doesn't.
A failed enumeration is a very large sample.
Any attempt to sample the same population will face the same difficulties in getting responses as a smaller sample. Otherwise, the methods of the small sample could be applied to the failing enumeration to get a better result set!
Trying to replace the enumeration with a sample is a bad idea, that is guaranteed to have an outcome that is worse, or no better, than the failed enumeration you want to replace.
A smaller sample is easier to get a better response rate from, dude.
This isn't hard stuff.
No, a smaller sample is not "easier to get a better response rate from".
That doesn't even make any sense. Why on Earth would you ever think that asking a question of fewer people would produce more answers than asking the same question the same way to more people?
"A smaller sample is easier to get a better response rate from"
Why? How so? Assuming you're sampling from the same population you're attempting to census, the response rate should be the same. If the response rate is different, there's a serious risk that the sample is non-representative of the population.
Smaller sample makes it easier to 1) concentrate your efforts to get a better count, 2) deal with systematic non-responses via weighting.
I'm not an expert, but you guys very much seem to be arguing from ignorance.
Source:
In a sample survey, counters pick particular blocks within census areas and ensure that everyone within them is counted. Given the laws of probability, the extrapolation of the estimates will be closer to reality than a universal count.
Google around; no one agrees with your simplistic take.
Are you seriously quoting a random guy on Quora? A self-described Canadian political science major? A guy that provides no evidence other than saying "the laws of probability"?
How about this: As a former PROFESSIONAL STATISTICIAN who WORKED at the Census Bureau, I tell you that a small sample will always produce a less accurate result than an even an incomplete enumeration using the same methodology. The error distribution of your sample will always be worse than that of the enumeration attempt.
Come on; you don't have to agree about the constitutionality of the issue to use common sense. How is it easier to get responses from every one of 1,000 people than every one of 320,000,000 people? That can't be a serious question.
If you are using the same methodology, and you are measuring response rate, your result likelihoods WILL BE IDENTICAL.
If you perform a large sample survey and a small sample survey using different methodologies, you can get different results. This is obvious. It is also not what we are talking about.
Again, ANY method you can use to increase the response rate in a small sample can also be applied to increase the response rate of a large sample.
Go ahead, try to come up with a counter example. Find me a method you can use to increase the response rate of your 1000 people that you cannot use to increase the response rate of 1001 people. Or 1500 people. Or 2000 people. Or 200,000 people. Or 320,000,000 people.
Okay: assign one census taker to each and every single person in your sample.
You can hire 1,000 census takers. You can't hire 320,000,000 census takers.
You most certainly can hire 320,000,000 census takers. There is no mathematical reason you cannot do so.
You'll still have a non-response problem, by the way.
As Torath says, you could indeed hire 320,000,000 census takers.
In fact, let's run the math. Let's pay each census taker 320,000,000 $10 an hour. Let's assign them each a single person. With a little care, you should be able to assign the census takers a person near them, who should, on average, take an hour or two to obtain their census subject. At only a single subject, we'll round up, and have each census taker paid for 2 hours.
That's 6.4 Billion dollars. Well under the cost of the 2010 census.
Sorry; despite all evidence from earlier in the discussion I thought you might be interested in speaking in good faith. I see that this was a mistaken assumption.
Sigh.
You don't even understand enough of the topic to rationally discuss it, and you accuse me of being unwilling to have a discussion in good faith?
You seem to be trying to make an argument about the efficiency of spending money to counter an argument about the statistical accuracy of sampling methods. In other words, rather than address the topic, you decide to pop off on something else entirely.
On top of that, you are just WRONG. I have no reason to treat your proposed "method" seriously, because you obviously don't mean it seriously. Despite that, I took the time to point out that yes, it is possible, and yes, the math behind a 320,000,000-large sample being better than a 1,000-large sample is STILL TRUE.
Also, in case you didn't notice, Armchair Lawyer pointed out that even at reasonable payment rates you can hire 320,000,000 'census takers' to each interview 1 person, and still fall within the current US Census budget.
If you want to have a serious discussion about the accuracy of sampling methods, then stick to it. If you want want to make a different argument, then admit that is what you are doing - and admit that you cannot argue against the fact that an enumeration is more accurate than a small sample!
That word "significant" is the keystone of your whole argument -- and, unfortunately, is both completely undefined and incapable of principled definition.
The simple fact is that every single question asked on a form provably reduces the response rate. And there is no way to extract a standard from the Constitution's text that would define what constitutes a "significant" undercount. Any standard allowing any questions other than number of residents would have to be made up arbitrarily by judges, reflecting not the actual Constitution but merely the judge's personal opinion of whether the question had an acceptably low burden.
The choices are very simple:
1) Enumeration only and no other questions, declaring that every single census ever taken was unconstitutional.
2) Whatever questions the political branches decide to ask.
3) Judicial review of each and every question asked on the Census as to whether it, alone or as a result of its cumulative effect, reduces the accuracy of the enumeration "significantly", in the opinion of the judge.
"The choices are very simple:"
It's a bit more complicated (or a lot simpler) depending on what constitutional question you ask.
Somin want's it to be an enumerated powers question. In which case, the effects on the accuracy of the enumeration aren't relevant and you only get options 1 and 2.
On the other hand, questions that might effect the accuracy of the enumeration could be individual challenged on equal protection or some other constitutional basis in which case the effects on the enumeration are relevant and can't readily be ignored. This approach takes options 1 and 2 both off the table.
No, it really isn't any more complicated, because the whole point is to reverse the dependency to clarify the issues. No matter what constitutional question you ask, you wind up with one of the three as the answer. Therefore, whichever of the three answers you choose dictates what constitutional questions you can ask.
Since Mr. Somin rejects both #1 and #2, he chooses #3. A perfectly logical argument from what look like reasonable principles is immediately and obviously exposed as a recipe for repeated acrimonious litigation with arbitrary results, when you work it from the result-up. Accordingly, the principles are unsound and need to be modified.
Liberals buy votes from unproductive people by promising them free stuff. That's the allure of sub-90 IQ third worlders.
Exactly why is the judiciary entitled to make a determination as to whether a particular census question that might discourage scofflaws from responding and thereby may produce a less accurate census that they deem to be unacceptable?
They are not. What they are entitled to do is determine whether the process of adding the question was carried in accordance with law.
It wasn't.
By your logic, a census that just makes up numbers would be fine, because it's not up to the judiciary to look into accuracy?
No, that would be in violation of the laws passed by Congress governing the taking of the Census, and the Courts would be the appropriate venue to deal with the violation of the law.
Similarly, a law passed by Congress ordering the Census Bureau to just make up numbers would be unconstitutional, as it would be in violation of the Constitution's requirement to perform the Enumeration of [people].
However, when Congress passes a law that tells the Census Bureau to perform such an Enumeration, it is not the place of the courts to decide that this methodology and all it's problems are OK, while that one has potential drawbacks that are unacceptable.
Plainly put, judges are nowhere close to qualified to answer that question. Politicians aren't either, but at least they answer to the People. The exact method used seems to be an excellent example of a political question, and the Courts should stay out of it.
Read the comment I'm responding to again.
I did? That's why I posted. Did you read my post?
Perseus says, and I agree, that it is not appropriate for the judiciary to make determinations about whether a particular methodology would produce an outcome they deem acceptable. They are not entitled to do so.
The only criteria they should be judging on is whether the methodology meets the basic requirements of the Constitution or implementing laws passed by Congress.
The comment argues that the courts should not look into the effects of administrative decisions on the accuracy of the census. This proves way too much, as I noted.
You've come in adding a whole new proviso about 'whether the methodology meets the basic requirements' which is a different thesis from above.
It is also wrong because, a) it ignores the ACA, and b) it begs the question about the citizenship inquiry meeting basic requirements for an accurate count.
Er, no, I don't believe that it is a different take on Perseus's answer, since it is generally an accepted provision that, unless otherwise stated, extreme absurdities are excluded from normal discussion. If you assume otherwise, I expect your next post to list all assumptions you are making first, before the rest of your argument.
As for the citizenship questing meeting the requirements, I'd say the almost 200 years of precedent settle that question thoroughly. Unless you want to claim that the Founders and 10+ generations of American jurists didn't actually know what they were doing... in which case, you're going to need a LOT more evidence.
And by ACA, I assume you mean the APA?
it is generally an accepted provision that, unless otherwise stated, extreme absurdities are excluded from normal discussion.
Don't be silly.
I'm sorry, your premise failed to account for the activities of the Lizard People, therefore you are wrong. All absurdities must be accounted for in normal discussion.
If non-citizens are supposed to be counted for the purposes of portioning representation, how do you explain the 3/5ths compromise? Half our founding fathers clearly didn't believe so because it was rife with abuse from the onset, thus the compromise to stop slave holders from literally buying and producing additional representation without a voting population.
"Congress has the power to take action X, but not if X maybe has a negative effect" is.... a creative way of coming at the question. That's like saying Congress has no power to declare a war they might lose.
[…] “Enumerated Powers and the Census Case,” by Ilya Somin […]